For nine months, a scientific expedition searched for blue whales in the waters around the Galápagos Islands but found none. Between October 2025 and July 2026, the researchers surveyed areas of the Ecuadorian archipelago where the giants had historically been recorded, hoping to attach satellite tags and learn more about the animals' migratory routes. The search ended without the team tagging a single whale.

The absence coincided with unusually warm ocean conditions during an El Niño event. Sea surface temperatures were as high as 82 degrees Fahrenheit (28 degrees Celsius), while at a depth of 330 feet (100 meters) water temperatures ranged from 66 to 67 F (18.9 to 19.4 C) — up to 9 F (5 C) warmer than usual for that time of the year.

This raised the possibility that El Niño changed where blue whales were finding food and, in turn, where they went. The researchers suspect the animals may be seeking colder, more productive waters. However, the whales themselves are still missing.

"The hypothesis is that the whales are staying in areas with greater productivity [more food] farther south, in waters with slightly cooler temperatures," Daniela Alarcón, the expedition's leader and a researcher at the Universidad San Francisco de Quito's Galápagos campus and the Galápagos Science Center, told Live Science.

As to where the missing whales have gone, possibilities include the waters off Peru and offshore areas where the animals could go undetected. In this scenario, the unusually warm waters around the Galápagos Islands would act as a "temperature barrier," Alarcón said, preventing them from feeding in the archipelago's waters.

Why water temperature matters

Blue whales (Balaenoptera musculus) can consume up to 12,100 pounds (5,500 kilograms) of krill — small crustaceans that make up the whales' primary food source — every day. That means they need to find areas with enough food to sustain them.

"These are very large animals that need to move to areas where they can keep feeding," Alarcón said.

It's not yet clear whether this year's El Niño caused a krill shortage in the waters off the Galápagos, but some indirect evidence points that way.

Past expeditions suggested that blue whales tend to feed in water layers around 71 F, which this season was identified around the Galápagos at a depth of about 197 feet (92 m). By contrast, peak chlorophyll concentrations, an indirect indicator of abundance of the krill's food source, phytoplankton ‪—‬ were recorded much higher in the water column, at between 39 and 121 feet (12 to 37 m) below the surface. Because krill abundance wasn't directly measured, that data suggests, but does not establish, that there was a mismatch between the whales' preferred feeding layer and the presence of their food.

"Honestly, we had never encountered such unusual conditions before," Juan Pablo Muñoz, a marine biologist at the Universidad San Francisco de Quito who was also on the expedition, told Live Science.

Continuation of a trend

The whales' absence did not come as a complete surprise to the team. Sightings had already begun to decline before the most recent expedition. Reports submitted through Cetácea Galápagos, a citizen science network, began to dwindle between August and September 2025. There were two isolated reports of blue whales, in May and June 2026, but the team could not confirm either one, despite intensifying its search.

"We knew conditions wouldn't be the same, but we thought that if we put in more effort, we might be able to find some individuals. But we searched, and searched, and searched — and found nothing," Alarcón recalled.

Part of that could be down to the elusive nature of the giant creatures. Humpback whales (Megaptera novaeangliae)are generally easier to spot, Muñoz noted. In some places, they can be seen from the shore or after a half-hour boat ride. Blue whales, by contrast, rarely approach mainland Ecuador's coast and tend to stay far from populated areas in the Galápagos.

Sensitive to temperature fluctuations

Fernando Félix, a marine biologist at the Pontifical Catholic University of Ecuador who was not part of the expedition, was unsurprised by the result.

"We already knew there wouldn't be any blue whales in the Galápagos," said Félix, who is also a researcher at the Whale Museum (Museo de Ballenas).

Félix told Live Science that there are historical precedents for this behavior in the southeastern Pacific. During the 1982-1983 El Niño, which also brought substantial ocean warming, blue whales and other species, including fin whales (Balaenoptera physalus) and Bryde's whales (Balaenoptera brydei), were recorded shifting southward.

"We might think that because they are 'just animals,' they lack memory or don't understand these processes — but we would be wrong," Félix said. "They can clearly sense that when temperatures are unusually high as they travel north from the south, vertical mixing has been disrupted and there is no food in the Galápagos. They perceive environmental conditions and naturally direct their movements toward areas where they believe feeding conditions are better."

Rather than indicating a mass die-off, the whales' disappearance may reflect a shift in their movements in response to changing ocean conditions, Félix said.

"I don't think the whales have died," Felix said. "They simply didn't go to the Galápagos region. They went elsewhere, where conditions were better."

This response to environmental conditions may also have impacted dolphin movements. Pods may have moved toward colder waters, where food could be more abundant, or may have even dispersed, Félix suggested.

When will blue whales return to the Galápagos?

El Niño's effects will likely persist over the coming months. "Between January and May 2027, we'll have warm water coming from the north and warm water coming from the west," Félix warned. "This is going to hit us extremely hard."

Blue whales, humpback whales and dolphins would likely return to Galápagos waters once El Niño subsides and ocean conditions begin to normalize. He said pelagic fish (species that live in the water column rather than on or near the seafloor) ‪—‬ including tuna, swordfish, anchovies and sardines, which have also moved toward colder waters ‪—‬ would return as well.

Muñoz agreed. Krill populations can rebound just one to three months after cold, nutrient-rich waters begin to rise to the surface. More whale sightings were recorded in the years following the 2015 and 2023 El Niño events. Based on these precedents, he suggested that 2028 could be a promising year if cold water returns to the surface. "The system bounces back," he said.

Félix compared El Niño to a computer restart. "It resets the environment, allowing ecological processes to begin again," he said. He's confident that blue whales will return to the area. "I can guarantee that there will be large numbers of blue whales at this time in 2028," he told Live Science.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Scientists in China are exploring what it would take to develop "the last AI built by humans" — a powerful, self-improving artificial intelligence (AI) system capable of creating its own successors.

In a new paper, researchers examine how approaches to recursive self-improvement (RSI) could help tackle the rising financial and computational costs of developing ever-more advanced AI models.

RSI refers to the idea that future AI models could direct their own training and development, using what they learn from each round of improvements to guide subsequent rounds of development.

This recursive feedback loop could eventually enable AI systems to develop newer, more powerful generations of models with minimal human input.

The study, which was uploaded to the preprint database arXiv Sept. 10 but has not been peer-reviewed, proposes a framework for classifying AI systems based on their self-improvement capabilities, from Level 1 (lowest) to Level 5 (highest).

At the lower levels, development remains largely human-led. As AI systems move up the scale, they become increasingly capable of identifying their weaknesses and deciding how to address them, eventually gaining the ability to make their own “meta-improvements” — improving how they improve themselves.

Is self-improving AI 'nonsense'?

Whether AI could ever achieve this level of sustained self-improvement remains an open question. Previous experiments have, however, demonstrated more limited forms of RSI.

In one study earlier this year, an AI system called A-Evolve-Training was tasked with improving the performance of a 30 billion-parameter large language model (LLM).

Over four rounds, A-Evolve-Training modestly improved the LLM's benchmark performance by testing different training techniques and using the results to change how it approached subsequent experiments.

Peter Bentley, a professor of digital biology at University College London, questioned whether such experiments offer a credible route to RSI.

"This idea that some AI is magically going to start generating better and better AIs by themselves [is] nonsense," Bentley told Live Science in an email.

Bentley pointed out that self-improving computer systems have been around for decades. Evolutionary computation and reinforcement learning, for instance, are both based on the principle of systems refining their performance over time in response to feedback.

The much greater challenge, he said, is developing AI systems capable of autonomously conducting the research needed to create more powerful successors — not to mention overcoming their dependence on human expertise and physical computing infrastructure.

"You can't 'recursively self-improve' unless you can also enhance your own training, enhance your hardware, and test it exhaustively in real world settings," Bentley added.

The challenge of true RSI

The study's authors acknowledged some of these challenges. Developing RSI-capable systems could require enormous investments in computing power, as well as human supervision to ensure successive improvements delivered meaningful gains. All of this becomes effectively redundant if the resources needed to develop and test subsequent AI models outweigh any benefits they offer, the authors noted.

Scientists would also need a reliable way to determine whether an AI system was genuinely becoming more capable over iterations, rather than exploiting weaknesses in tests used to measure its capabilities. This is a known problem in AI development, where models can achieve higher benchmark scores by exploiting testing methods or cherry-picking results.

Kayla Blomquist, an AI researcher at the University of Oxford and co-founder of the Oxford China Policy Lab, said that, while the paper doesn't provide a practical road map to RSI, it could help developers and policymakers understand the technical challenges and safety risks associated with increasingly autonomous AI systems.

She also questioned whether there would ever be a clear point at which AI became completely independent of human development. "The paper describes a gradual handover of responsibility rather than a single moment, so there may never be an obvious 'last' model built by humans," Blomquist added.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

More children, teens and very young adults are being diagnosed with colorectal cancer in the United States, and new research suggests they often experience aggressive forms of the disease.

While diagnosis rates are rising, the disease is still rare among young people. Colorectal cancer, which affects the lower intestines and rectum, is largely a disease of age: 55% of all colorectal cancers are diagnosed in people over 65, and only 16% occur in people ages 49 or younger, according to a 2026 study published in the flagship journal of the American Cancer Society. Only a few hundred cases per year are diagnosed in people under age 21 in the U.S.

Nevertheless, the numbers of cases in the very young are rising, even as colorectal cancer rates decline or hold steady in older individuals. This increase coincides with an overall rise in colon and rectal cancer rates in people under 50.

The reasons for this schism aren't entirely understood. People over 50 receive routine colonoscopies in which precancerous polyps are removed, and that likely accounts for much of the decline in colorectal cancer diagnoses in older adults. But researchers suspect that environmental factors, such as ultraprocessed foods and sedentary lifestyles, might be behind the growing rates in young people.

"Most of the research has really focused on the gut microbiome," the community of bacteria and other microbes that live in the gut, said Dr. Christopher Lieu, a professor of medical oncology at the University of Colorado Anschutz and a medical adviser to the nonprofit Colorectal Cancer Alliance.

Microbiome disruptions may spur inflammation, which, in turn, may trigger DNA mutations that cause cancer. But these mechanisms are complicated, Lieu told Live Science, and there likely isn't one simple reason young people's microbiomes may be getting disturbed.

"It's going to be extraordinarily more complex than that," he said.

Aggressive disease in young people

What researchers are starting to learn is that colorectal cancer in people in their early 20s or younger tends to be a more aggressive disease than that often seen in older people.

"We do feel like it is a bit of a distinct disease than what we're seeing in older adults," said Dr. Erica Arnold, a pediatric surgery research fellow at Nationwide Children's Hospital in Columbus, Ohio. In October, Arnold presented new research on colorectal cancer in people ages 21 or younger at the annual meeting of the American Academy of Pediatrics.

Arnold and her colleagues analyzed the cases of 84 patients diagnosed between 2012 and 2023. They found that the median age of diagnosis was 16 years old and the median duration of symptoms before diagnosis was only 12 weeks long.

Only about 20% of the patients had a close relative with colorectal cancer, suggesting that for many of these young people, the disease did not necessarily emerge due to hereditary factors.

Eighty percent of the study participants were diagnosed with Stage III or Stage IV disease, meaning the cancer had already spread locally and to lymph nodes or to other, distant organs when the patients were diagnosed. For comparison, in colorectal cancer patients of all ages, 60% are diagnosed in Stage III or IV, according to the Centers for Disease Control and Prevention (CDC).

In addition, Arnold said, one-third of the patients in their study had cellular markers of particularly aggressive cancer.

"Some of that might be related to delays in diagnosis," Lieu said, "but if you want my honest opinion, it does suggest that these cancers are developing differently."

The study's data was not nationally representative, but a recent analysis of CDC data has also found that colorectal cancer rates in youth are on the rise. In 1999, the diagnosis rate in kids ages 10 to 14 was 1 in a million. By 2020, it had risen to 6 in a million. For teens between 15 and 19, the rate of diagnosis went from 3 in a million to 13 in a million in that same time frame.

Unclear causes

There is no definitive reason for this rise, Lieu said. Research on the microbiome has suggested that processed food, antibiotic exposure in childhood, stress, reduced exercise and other lifestyle factors may contribute to this increased risk of colorectal cancer.

Some microbes may play a particularly strong role, at least in some cases of colorectal cancer. For instance, a specific variant of the common gut bacterium Escherichia coli is strongly associated with colorectal cancer; the variant carries a genetic segment and produces compounds that are toxic to genes. But this variant alone cannot explain all colorectal cancer cases in young people. Nor is it clear why the gut microbiome might be undergoing a shift in the first place.

"It may be a combination of food, antibiotic use and environmental exposures," Lieu suggested.

Regardless of the exact causes behind the cases, Arnold's new research pointed to the symptoms that deserve closer scrutiny in young people. Among the 84 patients in the study, 75% reported experiencing back and abdominal pain before their diagnosis, 45% had changes in their bowel movements, 43% experienced unexplained weight loss and 38% had rectal bleeding.

These are classic colorectal cancer symptoms, Arnold said, but they might be overlooked in this young age group because these same symptoms can have more common causes.

Parents should know that in most cases, stomach pain and gastrointestinal disruptions in kids are not caused by colorectal cancer, Lieu said, calling the cancer an "extraordinarily rare event." Nevertheless, pain or gastrointestinal upset that doesn't ease or that worsens over time needs a checkup, and symptoms like weight loss and blood in the stool are very unusual in kids and deserve extra attention.

"It's just important to investigate it to make sure we're catching colorectal cancer earlier rather than later in these patients," Arnold said.

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Actor Christina Applegate was diagnosed with multiple sclerosis (MS) over five years ago — but this week, she announced that a recent reevaluation found that she doesn't have the condition after all. In the wake of that revelation, she's been left without a diagnosis, and she said on a podcast that she felt she was "back at square one, where I don't know what's going on with me."

Applegate's case is not totally unique, although MS diagnosis has improved over the years. Studies offer different estimates of how frequently MS is misdiagnosed, with rates ranging from 5% to over 40%, depending on the patient population and clinical setting. There is also some degree of underdiagnosis — cases in which people with MS aren't flagged as having the disease.

So, how difficult is MS to identify, and what does that mean for patients? Live Science spoke with experts about the current state of MS diagnostics and the improvements that may be coming in the future.

"It's also really important that doctors and patients be aware that it's not inappropriate to revisit a diagnosis — to make sure it's right," Dr. Ari Green, chief of the Division of Neuroimmunology and Glial Biology at the University of California, San Francisco, told Live Science.

The importance of early diagnosis

MS is an autoimmune disorder in which the immune system attacks the central nervous system, meaning the brain and spinal cord. These attacks harm neurons directly and also take aim at myelin — a fatty, insulating substance that coats neurons' outgoing wires and helps them send signals efficiently.

This damage leads to distinctive scarring in nerve tissue and causes a range of symptoms, including vision problems, muscle weakness, tingling sensations, dizziness and bladder control issues. Patients with untreated and advanced MS can lose the ability to walk and stand or, in some cases, develop partial or complete paralysis.

These severe outcomes are becoming rarer as treatments have improved significantly, Green said. Current treatments keep immune attacks at bay and thus prevent further nerve damage.

"It's moved from people ending up with severe, profound disability at 10 or 20 years of disease, and now it's at like 30 or 40 years of disease, they have modest disability," Green told Live Science. The disease most often emerges between age 20 and 40, although it can occur in any age group. Importantly, Green said, "the earlier people initiate treatment who have MS, the better they do over the long term."

For that reason, it's important to reach an MS diagnosis efficiently, so that patients can start that critical treatment, said Dr. Yohance Allette, a neuroimmunologist and researcher at Penn State Health. "You're trying to be as proactive as possible with this."

"A diagnosis of exclusion"

However, there is no single, surefire test that can reliably show that a person has MS. "With MS, it comes down to the keyword 'criteria," Allette said.

MS experts have compiled a list of criteria, which are updated periodically, that reflect how the features of MS show up in lab tests; brain, optic nerve and spinal cord scans; and patients' symptoms. (Green was among the dozens of experts involved in the latest update, which was conducted in 2024 and published in 2025.)

So, what does diagnosis look like for a patient today? The first step is to recognize symptoms that might point to MS.

Those early symptoms — such as issues with balance, vision or bowel control — vary significantly from person to person and may not immediately raise red flags that the patient has a neurological problem, Green said. The way those symptoms emerge can be a clue, though; typically, MS symptoms worsen over the course of a day or a few days and then dissipate, he noted.

Once concerning symptoms are flagged, MRIs of the brain and spinal cord are taken. The scans may reveal evidence of myelin damage, and neurologists note the shape, size and location of those lesions. Damage often appears near blood vessels, since that's where immune cells access the brain, Green noted.

Other conditions — like migraine, uncontrolled diabetes or vascular disease — can also come with some myelin damage, so it's key to look for specific patterns related to MS. Neurologists who specialize in MS are practiced at this task, while doctors in other neurological specialties may not know how to distinguish MS-related myelin damage from that caused by other conditions, Allette and Green both noted.

MRIs of the optic nerves can also show MS-related damage, and these scans are emphasized in the latest diagnostic criteria, Allette noted. Electrophysiologic tests, which assess the function of nerves, are also "super important because they tell you something totally different than just seeing a lesion on MRI," Green said. Further signs of MS can be found via spinal taps, which sample the fluid around the brain and spinal cord that may carry proteins and inflammatory cells related to MS.

In addition to collecting these clues, doctors must rule out diseases that look similar to MS. These include neuromyelitis optica and myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) — two other autoimmune conditions that affect the central nervous system. Both these diseases are associated with distinctive, rogue immune proteins, called autoantibodies, that latch onto various parts of the nervous system; those specific antibodies are not seen in MS.

Other conditions — such as the inflammatory disease lupus and neuromuscular disease myasthenia gravis — can sometimes look like MS at first glance, as well. Some of these "MS mimics" can be ruled out with data from spinal taps or antibody tests, while others require a more laborious battery of assessments.

"Ultimately, it's a diagnosis of exclusion," Allette said. "There are things that you can rule out quickly," while others take time to investigate thoroughly.

Conditions like MOGAD and neurosarcoidosis, an inflammatory condition sometimes mistaken for MS, can actually be worsened by MS medications, Green noted. So if a patient's MS flare-ups worsen after they start a treatment, that could hint that something else is the culprit, he said. But because there may be long gaps between flare-ups, doctors may need to track a patient for some time before their worsening becomes apparent.

Gaps to fill?

Given all the tools available, a patient can be reasonably confident in their MS diagnosis, Green said.

"In general, there's a fair amount of diagnostic confidence that comes with MS diagnosis today," he said. However, that confidence is highest when the tools are in the hands of a specialist who is following the most up-to-date criteria, he noted. The test results require interpretation, and that's one place where uncertainty can creep in.

Regarding Applegate's circumstances, "to speak bluntly, I think the case came as a bit of a shock because we always expect that part of society to have quick access [to high-quality medical care]," Allette said. For some patients, the road to accessing specialty MS care can be long, which can result in missteps, and even misdiagnoses, along the way. Applegate went five years with an incorrect diagnosis, but "would that have been seven or eight years for somebody who doesn't have that access?" Allette speculated.

Without knowing the details of Applegate's case, Allette and Green said it would be difficult to speculate on what resulted in her misdiagnosis. (Applegate mentioned that her brain scans had been reassessed with a "new technology," but didn't go into specifics.)

Looking toward the future of MS diagnostics, Green and Allette think emerging proteomic tests hold a lot of promise. Such tests generate a profile of the proteins and smaller fragments, called peptides, in a patient's blood or cerebrospinal fluid. They can be designed to look for proteins involved in driving the underlying immune attacks, as well as signs of nerve damage that MS wrecks, they explained.

These in-depth, protein analyses could help paint a more comprehensive picture of MS, which could ultimately help scientists "look at that whole system in concert" to reach a strong diagnosis, Allette said.

No proteomics-based diagnostic tests for MS have been approved yet, Green clarified. And even with those upcoming advances, Green doubts there will ever be a simple, one-time test that can determine a patient has MS. "I think that's probably unlikely, that it would ever be that straightforward," he said.

He does think that, in time, scientists could pare down the number of tests required to reach a diagnosis, but he argued that the clinical evaluation conducted by a neurologist will likely remain crucial. "I'm doubtful that there's going to be a test that will fully replace that," he said, "at least in the foreseeable future."

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

With artificial intelligence (AI), scientists have a powerful new tool for creating never-before-seen proteins, which can be used to make either groundbreaking new medicines or devastating bioweapons.

To help guard against nefarious uses of these proteins, scientists have recently figured out a way to "watermark" them without compromising their function. A team from Google DeepMind developed this technique to better track the provenance of AI-designed proteins.

"This has the potential to reduce biosecurity threats, such as unauthorized toxin creation, and may help with database curation, preventing artificial designs from cluttering the standard databases," said Roy Sleator, a biology professor at Munster Technological University in Ireland, who was not involved in the research.

"In practice, it could help trace a design's origin, flag engineered proteins for extra biosecurity checks, and reduce misleading claims about how a protein was made," Sleator told Live Science in an email.

However, in its current form, the same AI technologies that add watermarks can also strip them away, other experts warned.

New ways to hide watermarks

The unveiling of the AI program AlphaFold in 2020 was a milestone in protein engineering. The tool, which won Demis Hassabis, John Jumper and David Baker the Nobel Prize in Chemistry in 2024, can reveal the 3D structure of a protein from the sequence of its building blocks, called amino acids. Proteins fold like origami, adopting 3D shapes that dictate their function.

However, AlphaFold's technology also opened the door to generating brand-new proteins that could be used to create biothreats, including viral proteins that can evade our immune systems and existing vaccines.

"Anything that can be used for good can be used for bad," said George Rose, professor emeritus of biophysics at Johns Hopkins University.

Current screening tools identify novel dangerous or restricted molecules by seeing how similar their amino acid sequences are to those of recognized threats. But the problem is that AI could come up with proteins that don't look like anything we've ever seen. That's what the new watermarking technique aims to solve.

The DeepMind team has framed the watermarking effort as a "family" of related techniques they call SynthIDBio. The first member of this family is SynthIDBio-sequence, which encodes watermarks into the amino acid sequences of proteins created by a program called ProteinMPNN.

When users tell ProteinMPNN the desired 3D shape for their protein, it chooses amino acid options that ultimately fold into that shape. SynthIDBio-sequence subtly influences how the program chooses each amino acid according to a secret watermarking "key." Once the sequence is generated, anyone who knows the hidden code can assess how likely the sequence is to have been created under the key's bias. In their recent study, the DeepMind team showed that proteins built with this watermark could still bind to their targets as well as non-watermarked proteins.

A sister technique, called SynthIDBio-structure, instead embeds a watermark directly into the protein's structure. It does this by altering the function of AlphaFold 3, DeepMind's flagship protein prediction algorithm.

When AlphaFold predicts a structure, SynthIDBio-structure gives it a side mission: The AI must make minuscule tweaks to how the atoms in the protein are arranged. These tweaks can be picked up by a detector later, and they're small enough that they don't look like obvious distortion.

In a test, these two systems created hidden watermarks that could be detected with incredible accuracy, flagging more than 99% of watermarked proteins and correctly ignoring 99.9% of non-watermarked proteins.

How to bypass a protein watermark

Live Science spoke to several experts who agreed that SynthIDBio was an impressive technical feat. However, they said it wasn't ready for widespread implementation yet.

The two techniques are "effective but brittle," Bruce Wittmann, a principal applied scientist at Microsoft, told Live Science in an email. That's because, as the DeepMind team also notes in their paper, the watermarks can be easily removed.

"For sequences, watermarks can be removed by 'resequencing' — effectively, altering the protein amino acid sequence such that the watermark is no longer detectable," Wittmann said. "For structures, watermarks can be removed by adding noise to or otherwise modifying the spatial coordinates of the structure." Screening tools wouldn't be able to detect if a watermark had been removed from a protein.

Experts are also skeptical that watermarking would provide a long-term solution for biosecurity.

An indelible watermark could identify who originally made a nefarious protein, or identify when a user was accessing a protein generation software without permission. This might deter some criminals, but terrorists looking to create a novel protein for a biological weapon would be less likely to care about being identified after a successful attack, said Max Brackmann, a chemical, biological, radiological, and nuclear arms expert at the Spiez Laboratory in Switzerland.

Additionally, the vast majority of software for protein prediction and generation is open-source, meaning much of the code is freely available to use and modify. That makes it far harder to bake watermarking into these programs. Wittmann warned that the field would have to shift toward closed-source models before watermarks could become an industry standard.

He added that watermarking could act as one layer of a biosecurity defense shield, rather than offering standalone protection. Top-down regulation would be challenging to enforce. "I imagine it would be very difficult to ensure the regulations were being followed," he said.

Rose was even blunter about regulating the use of generative proteins.

"I'm convinced it won't work," he said. "Once there's a sweet technology that does something, government regulation to keep it under wraps has never been really successful."

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

On Monday (Oct. 19), Japan's space agency launched a first-of-its-kind sample-return mission to Mars. The historic endeavor marks another major step toward humanity's shared goal of understanding — and perhaps eventually colonizing — one of our closest planetary neighbors, experts say.

The Japan Aerospace Exploration Agency's (JAXA) Martian Moons Exploration (MMX) probe is an international effort, constructed with help from NASA, the European Space Agency (ESA) and others. Over the next few years, it will closely study the Red Planet's twin moons, Phobos and Deimos. The spacecraft will snap close-up photos of the pair before attempting to land on the larger satellite, Phobos, to scoop up rocky material that will hopefully be carried back to Earth in around five years.

If the mission is successful, the returned samples will finally settle a decades-old mystery about the origins of both moons. The findings could also help paint a picture of Mars' ancient history and provide clues about what the early solar system was like.

"This mission marks a transformative leap in our capacity to investigate the Martian system," Sonia Fornasier, an astronomer at the Paris Observatory who has studied Phobos and Deimos extensively and is a project scientist for MMX's infrared spectrometer instrument (MIRS), told Live Science. It "ranks very high" among the missions we have sent to the Red Planet so far, she added.

What will MMX do?

JAXA launched the MMX probe on board an H3 rocket, which lifted off from Tanegashima Space Center in Kagoshima prefecture at 3:41 p.m. EDT Monday (4:41 a.m. Tuesday local time), Live Science's sister site Space.com reported. (The mission was initially scheduled to launch in 2024, but it was delayed due to initial concerns about the H3 rocket.)

The camper-van-sized spacecraft successfully separated from the rocket's fairing just over an hour later, and on Wednesday (Oct. 21), JAXA announced that the mission had passed the "initial critical operations phase," meaning the probe is now locked into its roughly 140 million-mile-long (225 million kilometers) journey to Mars, which will take around 10 months.

"MMX took its first step today," Naoki Murakami, general manager of Mitsubishi Electric’s space division, which led the design of the MMX probe, said in a postlaunch news conference early Tuesday morning (Oct. 20), according to The Japan Times. "This marks the beginning of a long, five-year journey."

Once the probe arrives at the Red Planet in August 2027, it will insert itself into a "quasi-satellite orbit" around Phobos. In other words, MMX will circle Mars alongside the moon but will not be gravitationally bound to it, similar to around half a dozen "quasi-moons" that currently accompany Earth around the sun.

After collecting data from a distance for a few months, MMX will swoop down and attempt to land on the surface of Phobos ‪—‬ not once, but twice.

During the first landing, MMX will hover just above the moon and deploy a mini rover, dubbed Idefix, onto the moon's surface. The wandering robot, built by the French and German space agencies, will have to contend with extremely weak gravity as it trundles across the surface of Phobos: The rover weighs around 55 pounds (25 kilograms) on Earth but will weigh only around half an ounce (15 grams) on Phobos, according to ESA.

Once Idefix has completed its initial assessment of the moon's surface, the entire MMX probe will set down on Phobos, on top of the spacecraft's four-legged lander module. Mission scientists hope this goes a lot more smoothly than the attempted landings of Japan's "Moon Sniper" and Resilience lunar landers, which crash-landed on Earth's moon in 2024 and 2025, respectively.

Two of the MMX lander's legs contain built-in sampling devices ‪—‬ one that will drill into Phobos and another that will hoover up surface material. The second device, dubbed the Pneumatic Sampler (or P-Sampler), was built by the Blue Origin-owned private company Honeybee Robotics on behalf of NASA and will collect its material in a "matter of seconds," Kris Zacny, vice president of the Exploration Technology Group at Honeybee Robotics, said at an Oct. 15 NASA news conference. "You blink, and it's all done."

In total, JAXA aims to collect around a third of an ounce (10 grams) of material from Phobos. That would be around twice as much material as the space agency managed to snag on its most recent sample-return mission, Hayabusa2, which collected samples from the "spinning-top" asteroid Ryugu in 2019, finding materials rich in surprising organic compounds.

Once the samples have been acquired, at some point between late 2028 and early 2029, the probe will conduct a series of close flybys of Mars' smaller moon, Deimos, before beginning the long trip back to Earth in late 2030 — the first return flight ever attempted by a spacecraft sent to Mars.

When the probe returns to Earth in 2031, it will release its samples in a 20-inch-wide (60 centimeters) container that will fall back to our planet's surface, most likely somewhere in Australia, according to ESA.

What could we learn about Phobos and Deimos?

Mars' twin satellites, which were first spotted by American astronomer Asaph Hall in 1877, are nothing like Earth's constant companion. For starters, they are much smaller; Phobos is around 14 miles (22 km) across, while Deimos is just 7.5 miles (12 km) wide. The duo also have lopsided shapes and are between 12 and 64 times closer to Mars than the moon is to Earth. (Phobos, meaning "fear," and Deimos, meaning "terror," are named after the sons of the Greek war god Ares, the counterpart to the Roman war god Mars.)

Phobos is dominated by the 6-mile-wide (10 km) Stickney crater, which was left by an ancient asteroid impact. The larger moon is also slowly falling toward Mars and may eventually be ripped apart before it reaches the Red Planet's surface.

Other Martian spacecraft, such as the Emirates Mars Mission orbiter, have flown close to the moons, and NASA's Mars rovers have frequently looked up to see the satellites eclipsing the sun in Mars' sky (sometimes both at the same time). Yet many mysteries remain about the pair.

"Mars is right in our cosmic backyard, and yet we haven't explored the planet's moons," Nancy Chabot, a space exploration expert at the Johns Hopkins Applied Physics Laboratory in Maryland and the deputy principal investigator of NASA's MEGANE instrument on board MMX, told Live Science in an email. "MMX will do this!"

The biggest mystery surrounding Phobos and Deimos is their origin. Some researchers suspect the satellites are asteroids, or maybe even comets, that have been captured by Mars' gravity, while others think the moons emerged from the debris of an ancient impact on the planet's surface, similar to how Earth's moon formed.

Neither theory is conclusive, meaning the moons' origins "remain enigmatic," Christopher Edwards, a planetary scientist at Northern Arizona University and one of NASA's participating scientists for MMX, told Live Science.

For example, explaining how Mars captured the moons is "quite hard from an orbital mechanics perspective," but if they did form from a debris disk from an ancient impact, they should have coalesced into better-defined spheres, Edwards said.

"MMX will solve this debate thanks to the spacecraft observations and the detailed analysis of the samples of Phobos on Earth," Fornasier said. The data the probe collects from its flybys of Deimos will also allow researchers to generate a "full characterization of this body," she added.

Fornasier is "particularly excited" by the possibility that the moons are captured asteroids because that would mean they would likely contain "organic and hydrated materials," which would have a big "astrobiological implication" and would shed light on what the early solar system was like.

What could we learn about Mars itself?

Regardless of how the twin moons formed, researchers also stand to learn a lot about Mars and how planets form and evolve.

"Phobos and Deimos are interesting, but not just in their own right," Tom Statler, an MMX program executive and scientist in NASA's Planetary Science Division, said at the Oct. 15 news conference. "Their presence is a key piece of the larger story of the Mars system. By uncovering the moons' origin, we'll gain new insights into the history of both Mars and Earth."

If the moons did form via an impact scenario, the samples MMX collects may also contain actual pieces of Mars for researchers to study.

"If these objects are in fact pieces from the Martian surface that have coalesced moons, we'll collect the best-preserved samples of the Martian surface," Edwards said. Other fragments of Mars have already reached Earth as meteorites. However, entry into Earth's atmosphere can transform them.

If Phobos is a chunk of Mars, that would be a real boon to NASA, which canceled its Mars Sample Return program — a mission that would have retrieved samples collected by the Perseverance rover — earlier this year. The agency's researchers will receive up to 15% of the MMX samples to analyze.

During its three-year stay in the Red Planet's system, the Japanese probe will also collect lots of data on Mars that could contribute to a wide range of future findings. "MMX will conduct in-depth studies of the Martian atmosphere, with a particular emphasis on dust and water transport mechanisms and the escape processes shaping Mars' atmospheric dynamics," Fornasier said.

This atmospheric data could temporarily fill a gap left by the recent death of NASA's iconic MAVEN spacecraft, which stopped sending data back to Earth late last year.

Analysis of craters on the Martian moons could also provide clues about other events in Mars' history. For example, some researchers think the Red Planet may have once had rings like Saturn or a larger moon.

Other experts are hopeful that the MMX findings could help lay the groundwork for future missions to explore ‪—‬ and perhaps colonize ‪—‬ Mars. "The samples from MMX will bring forth paradigm-shifting science on the Mars system for generations to come, especially as NASA prepares to send the first humans to Mars," Nicky Fox, associate administrator for NASA's Science Mission Directorate, said in a recent statement.

The next sample-return trip to Mars will be China's Tianwen-3 mission, which is scheduled to launch in 2028 and will aim to land on and retrieve samples directly from the Red Planet's surface, according to Space.com.

Tesla has unveiled a new "flying" version of its Roadster electric vehicle (EV) that uses SpaceX tech to pull off seemingly impressive airborne acrobatics.

During a filmed demonstration at the SpaceX rocket testing facility in Texas Thursday (Oct. 15), the second-generation Tesla Roadster was shown accelerating up a curved ramp before launching itself into the air, performing a 180-degree barrel roll and touching back down to Earth.

In another stunt, the vehicle leapt over a row of parked Teslas. For safety reasons, the vehicle was operated remotely in both instances.

"Turns out, the future does have flying cars," Tesla and SpaceX CEO Elon Musk wrote on X following the demonstration.

The Roadster's airborne capabilities are made possible by a system of cold-gas thrusters developed using technology from Musk's aerospace company, SpaceX. These thrusters release bursts of high-pressure gas to generate thrust, allowing the vehicle to accelerate rapidly and, according to Tesla, become airborne for short periods.

The new Roadster will come in three variants: Performance, Thrust and Rocket. Only the Rocket is capable of flight. The Performance model is a two-seater equipped with a suction system that's designed to improve traction and handling (or drive up inverted ramps, as in Tesla's demo), while the Thrust model adds rear-mounted thrusters that Tesla claims can propel the vehicle from 0 to 60 mph (97 km/h) in under a second.

The highest-end model, Rocket, features SpaceX cold-gas thrusters positioned around the vehicle to enable short bursts of flight and in-air control. This single-seat model will be available by invitation only, according to Tesla.

On the company's website, Tesla representatives explicitly stated that the Rocket isn't approved for use on public roads and that availability depends on regional regulations. During the reveal, Musk told the audience that the Rocket "is not necessarily a safety-optimized car."

Tesla hasn't disclosed how far or for how long its EV can remain airborne, nor is it the first company to develop a hybrid road-air vehicle. There is also skepticism from outside viewers on whether the demonstration could be replicated outside of controlled scenarios.

Artificial intelligence (AI) scientist Yann LeCun ‪— a professor of data science, computer science, neural science and electrical engineering at New York University ‪— described the demonstration on X as a "marketing stunt" and questioned whether the vehicle would work in practice.

"Even assuming that 50% of the mass of the car is the compressed air, that gives you less than 10s of useful thrust (with thrust-to-weight ratio larger than 1.0). Probably more like 5 seconds, actually," he posted on the social media platform. "So the car must be an empty, lightweight shell containing big compressed air tanks. Cute demo, but totally impractical. But of course, you knew that."

He likened it to SpaceShipOne, which debuted in 2004 as the first privately funded spacecraft and made three record-setting flights before being retired.

California-based Alef Aeronautics, which is developing its own flying EVs, marked Tesla's unveiling by releasing footage of its vehicle flying over an earlier Roadster model. Alef representatives said the demonstration marked its 1,000th logged flight.

"Welcome Tesla to the flying car race, we've been here alone for eleven years," Alef CEO Jim Dukhovny said in a statement emailed to Live Science. "Having competition is a good thing, especially from such a major company as Tesla."

Unlike Tesla's compressed-gas tech, Alef's vehicle uses electric propellers for vertical takeoff and landing, with a rotating cabin and body designed to support forward flight.

Production of the new Roadster is scheduled for 2028, but Tesla doesn't have the best track record of meeting its deadlines. The company has delayed production of its Optimus humanoid robot a few times since first unveiling the machine in 2021, for example, with an initial start in 2023, before being pushed back. Only in recent months has production begun to scale to a few hundred units per week.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Rising temperatures around the globe are contributing to a range of health risks, from infectious disease to food insecurity and wildfire smoke. Extreme heat is another major public health concern, as climate change drives up both average temperatures and the frequency of extremely hot days. This August was the world's hottest on record, and it drove upticks in heat-related emergency visits and deaths in various U.S. states.

Hotter weather is expected to pose dangerous health risks for millions, and pregnant people are especially vulnerable to its impacts, experts caution.

"Pregnant people are physiologically more vulnerable to heat," Dr. Susanna Esposito, a pediatrician at the University of Parma in Italy, told Live Science. "Climate adaptation should treat pregnant people as a priority population."

Extreme heat is linked to adverse birth outcomes that can be fatal for both mothers and newborns. Heat can also hinder fetuses' development, contributing to long-term health and cognitive effects.

Scientists and public health experts are working to identify the thresholds between safe and dangerous heat for maternal and prenatal health. As the planet warms, how will that affect the risks associated with pregnancy? And how can those risks be countered?

Low birth weight

When the body tries to cool down, blood moves to the extremities to dissipate heat through the skin's surface. Some studies suggest that, in a pregnant person, this may reduce blood flow to the placenta, which in turn reduces the organ's supply of oxygen and nutrients. This is thought to contribute to low birth weight, which is associated with a heightened risk of illness and death shortly after birth, as well as cognitive and mental-health challenges later in life.

Heat stress may also spike the body's levels of cortisol beyond what they should be in a healthy pregnancy, studies in lab animals suggest. This may expose developing fetuses to more cortisol than usual, affecting some aspects of development and functions of the placenta. This mechanism is not fully understood in humans yet, though.

That said, several studies in humans support the notion of a link between heat exposure and low birth weight. Mothers who were pregnant during a 2003 heat wave in France were more than twice as likely to have a baby of low birth weight than mothers whose pregnancies did not coincide with the extreme heat. The trend was especially strong when the heat wave hit during a person's first trimester.

In a study conducted in Arizona, hotter temperatures during the third trimester were also associated with lower birth weight, said study co-author Kristin Mickelson, a social-health psychologist who studies public health at the Arizona State University.

Preterm and early birth

Preterm delivery — before 37 weeks of gestation — is one of the most concerning health outcomes that's been linked to heat, Esposito told Live Science. The risk of neurological and developmental disabilities is higher in preterm babies than in full-term babies.

"If you're born preterm, you have a heightened risk of all sorts of adversities throughout infancy, but then also extending into childhood and even longer term," said Clara Heil, health researcher at Lund University in Sweden.

Severe heat may make preterm birth more likely. A review of 70 studies suggests that the chance of preterm birth is 16% higher when a multiday heat wave hits during the final weeks of pregnancy. Regarding average temperatures across an entire pregnancy, each additional 1.8 degrees Fahrenheit (1 degree Celsius) boosts the risk of preterm birth by 5%, the same review found.

High temperatures have also been tied to early births, which occur after 37 weeks of pregnancy but before 39 completed weeks. Early birth is associated with an increased risk of hospitalizations through childhood and slower neurologic development.

A study analyzing 56 million U.S. births across two decades suggested there was about 1 more early birth per 100,000 pregnancies on days with a maximum temperature above 90 F (32.2 C), compared to 60 to 70 F (15 to 21 C). Deliveries on hot days were typically about six days earlier than anticipated, but some were up to two weeks early.

With average temperatures continuing to rise, early and premature births could become more common. The United Nations predicts that the world is set to cross the 1.5 C (2.7 F) threshold of warming compared with pre-industrial levels, likely within the next five years, and we may even surpass the 2 C (3.6 F) threshold in that timeframe.

"We can probably expect to see these temperatures, or worse, in the upcoming years," Lizbeth Burgos Ochoa, statistician at Tilburg University in the Netherlands, told Live Science.

Preeclampsia, heart defects and other risks

Maternal heat exposure has also been associated with an increased risk of preeclampsia, a pregnancy complication related to blood pressure that can be fatal. In a 2011 study conducted in France, hotter temperatures during the first trimester were associated with a slightly increased risk of severe preeclampsia, although more research is needed to understand why.

Heat exposure has also been linked to other conditions of pregnancy, including gestational diabetes and hypertension, as well as increased emergency hospital admissions during pregnancy, overall.

Additional studies have highlighted potential risks for developing fetuses and newborns. For example, following a heat wave in Canada, the risk of a baby being born with critical heart defects was 6% higher, one study found. Babies with the highest risk had been exposed to temperatures of 86 F (30 C) or higher for at least 15 days during the first trimester of pregnancy.

Extreme heat waves may also raise the chance of stillbirth, according to a study conducted across six U.S. states. Four consecutive unusually hot days — exceeding the 97.5th percentile of temperature for a given county — in the last week of pregnancy was linked to a 3% increase in the risk of stillbirth.

These studies are observational, so they show links between heat and pregnancy outcomes but cannot directly show that the heat spurred the increased risks. They also don't take into account study participants' home environments or strategies for cooling off, so they don't provide a full picture of how much heat they were exposed to.

Caring for pregnant people and the next generation

Researchers are still working to understand what temperatures pregnant people should be most concerned about.

"It would be [helpful] to understand what level of heat becomes dangerous to women," Heil said. "Is there a threshold where it becomes difficult for the body to dissipate heat? And especially for the pregnant body, which is different from the non-pregnant body, to cope with that heat?"

Heil and her research team synthesized over 100 studies on heat and poor birth outcomes, such as stillbirth, preterm birth, placental abruption and preeclampsia. The studies varied considerably in that they covered different timeframes, focusing on just one trimester or all three, for instance. But in general, their results converged on a pattern: The risk of poor outcomes rose as temperatures hit about 87 F (30.5 C). At 102 F (39 C), the combined risk of poor outcomes rose by 20%.

This summer, daytime maximum temperatures in the continental U.S. averaged a record-breaking 88.6 F (31.4 C). The number of days per year over 105 F (35 C) is expected to triple in the U.S. by midcentury.

Heat's detrimental effects on birth outcomes are even more extreme for those with disadvantaged socioeconomic conditions, Burgos Ochoa noted. People with lower incomes are less likely to have air conditioning, and neighborhoods with a lower median income tend to have less green space, which contributes to trapped heat in their immediate surroundings. Expanding green spaces and adding bodies of water can help cool down neighborhoods and mitigate heat impacts on health, she said.

Heil's research team has developed a phone app for pregnant people and those who have recently given birth that forecasts the risk of heat stress and suggests actions to take to mitigate those risks, such as hydrating or changing clothing. Heil is hopeful that the app and similar early-warning systems can make a difference.

"Maternal health should be at the top [of our concerns], because it is about the next generation," Mickelson said. "If we don't take care of the mothers that are giving birth to the next generation, that's going to have long-term consequences down the road."

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

If you're approaching menopause, you are likely being bombarded by online ads for treatments promising relief for everything from hot flashes to brain fog. Some of these therapies have been approved by the Food and Drug Administration (FDA) and have strong evidence to support them — but one hormone-based treatment that's gaining in popularity has much less science to back its use for menopause.

More doctors are writing their menopausal patients prescriptions for testosterone therapy. A 2026 study conducted in the U.S. recorded over 90,000 off-label prescriptions for testosterone to women in 2025, mainly among women ages 45 to 64 and those with diagnoses of menopause. This marked a 58.7% increase from 2024.

These prescriptions are "off-label" because the FDA has not yet approved testosterone as a treatment for menopause, perimenopause (the transition into menopause) or postmenopause (the time period afterward).

So, does scientific evidence support the use of testosterone to alleviate menopause symptoms? Or is it too soon to say?

No direct link to menopause symptoms

Although often called a "male" hormone, testosterone is also produced by females, albeit at lower levels than in males, on average.

There's a common misconception that females experience "androgen deficiency" — a marked decline in testosterone and related hormones — leading up to menopause, said Susan Davis, a professor of endocrinology and the head of the Women's Health Research Program at Monash University in Australia. This is not the case.

Rather, in females, testosterone levels tend to drop slowly between ages 20 and 60, and there's a slight rebound from age 60 onward. Menopause, which generally occurs around age 51 to 52, is not linked to a marked decline in testosterone. Estrogen and progesterone fluctuate during perimenopause and then decline sharply, and these changes clearly contribute to the telltale symptoms of perimenopause and menopause.

Placebo-controlled trials have not been able to show that testosterone treatment helps to alleviate brain fog or improve well-being, mood, bone density or muscle function in menopausal patients, said Davis, who has conducted a number of these trials herself.

"Telling women that fatigue, low mood, brain fog are symptoms of androgen deficiency is completely untrue," she emphasized.

On Sept. 17, the FDA held a public workshop to evaluate the evidence for testosterone therapy for menopause, concluding that more evidence is needed to fully assess the use of the hormone as a menopause treatment.

Testosterone for low sex drive?

There is some evidence to suggest that testosterone might help to treat a condition called hypoactive sexual desire disorder (HSDD) in postmenopausal women. HSDD describes when a person has little to no sex drive and it causes them distress. Men and women of many ages experience it, so the condition is not unique to menopause. But it can sometimes develop in perimenopausal and postmenopausal women.

A review of 36 randomized trials involving 8,480 participants found that testosterone treatment delivered with skin patches was associated with 0.85 additional satisfying sexual events per month in postmenopausal women, compared to a placebo. This observed effect is not huge, Dr. Nanette Santoro, a professor of obstetrics and gynecology at the University of Colorado School of Medicine, told Live Science in an email.

Lately, anecdotal accounts of postmenopausal women having "fantastic sexual awakenings" after taking testosterone have been floating around, Santoro said. But these anecdotes haven't been backed by hard data yet. "Patients are being bombarded with a tsunami of information providing them with the most sky-high expectations imaginable," she said.

Notably, testosterone is not FDA approved as a treatment for HSDD in menopause yet. A drug called flibanserin, which affects serotonin in the brain, is approved to treat HSDD for pre- and postmenopausal women under age 65.

Placebo effects muddy the data

The placebo effect — when patients perceive an improvement in their health or well-being despite receiving a treatment with no active ingredient — can also complicate testosterone trials.

Davis said that in her patient groups, it's common for about 1 in 3 trial participants to report an improvement of their symptoms even though they are taking a placebo. Other trials have reported similar findings in their placebo groups, with about 30% to 40% experiencing some improvement in sexual desire and satisfaction, for example.

"I wouldn't say that there is definitely not a benefit," Davis said. "It's just that a benefit has not been shown."

One study trying to address this gap is the U.K. National Institute for Health and Care Research's ESTEEM trial. This large-scale, gold-standard trial aims to recruit hundreds of menopausal women and test whether using testosterone cream can reliably improve their quality of life. The cream will deliver testosterone through the skin and be applied to the lower abdomen and thighs. (Doctors often advise women to apply the creams in this way because those body parts are easy to hide if the treatment causes localized hair growth that the patients would prefer to conceal.)

The trial runners will track participants' symptoms linked to menopause, looking at sleep quality, memory performance, feelings of anxiousness, migraines, body mass, and overall well-being. The results of that trial are currently expected by 2028.

In the meantime, though, there isn't strong evidence that testosterone directly improves symptoms of menopause beyond HSDD. "Of all the studies done to date, if there was a screaming benefit, you think you would have seen it," Davis said.

Could taking testosterone for menopause be harmful?

Testosterone treatments available in most countries are not necessarily dosed for use in female patients. Only four countries — Australia, South Africa, New Zealand and the United Kingdom — have authorized a testosterone cream that's specifically formulated and dosed for women. These products differ from testosterone therapies for men in that they contain lower doses of the hormone. They are available to postmenopausal women with HSDD, specifically.

In other countries, such as the U.S., doctors are left trying to adapt medication dosed for males for their female patients, which can lead to a wide variation in dosing. And the effects of these different dosing regimens are poorly understood for female patients. "It is not necessarily unsafe to use male preparations for women, it's just that much, much less must be used," Santoro said.

Excessive doses raise concern about unwanted masculinizing effects, such as clitoral enlargement, voice deepening, hair thinning, baldness, and rapid growth of facial and body hair, Santoro said. (These masculinizing effects might be sought after in the context of gender-affirming care for transgender men and transmasculine people, but would often be undesirable for cisgender women.)

Additionally, the potential impacts of extra testosterone on cardiovascular health and breast cancer risk are poorly understood, Santoro noted. At lower doses, testosterone can cause mild acne, oily skin, and some increase in facial and body hair growth in postmenopausal women, trials have found; if these symptoms remain mild, they are generally considered acceptable.

Testosterone is being used off-label in the U.S., which means the "safety, efficacy and appropriate dosing for that use have not been evaluated and approved by the FDA," a spokesperson for the agency told Live Science in an email. The dosages and formulations approved for males "may differ substantially from what might be appropriate for women, whose testosterone levels and metabolic profiles are clinically distinct," the spokesperson said.

Anecdotally, Santoro encountered one female patient who had received very high levels of testosterone through an implant placed under the skin, which had been prescribed to her by a different doctor. She started losing her hair and seeing sizable growth in her clitoris, while her breasts started "shrinking away."

"Women are being prescribed some very big doses in North America, and no one studied the safety of these," Santoro said.

At the end of the day, Davis argues that women deserve to be given accurate information, given that navigating menopause can already feel like "swimming through wet cement." Good menopause care not only involves discussing potential pharmacological treatments, but also evaluating a patient's lifestyle, such as diet and exercise, and discussing societal pressures that often affect menopausal people. For instance, they may face pressure to maintain an active sex life and younger-looking physique.

When it comes to hyped therapies like testosterone, she said, "just be cautious about people who tell you they've got the answer that no one else has got."

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Since June, more than 300 southern right whale (Eubalaena australis) calves have died in Argentine Patagonia, in what scientists describe as an "exceptional" die-off. The preliminary figure already far exceeds the previous annual record set in 2012, when 113 southern right whale calves died — the highest number ever documented for the species, both locally and globally.

Now, scientists are racing to figure out exactly what's causing the record-breaking die-off.

Laboratory analyses of specimens that died this year tested positive for the pathogenic bacterium Erysipelothrix rhusiopathiae, environmental authorities in Chubut province told Live Science. E. rhusiopathiae was first identified in southern right whale calves in 2013, but scientists don't know what might have changed this year to cause an infection-driven die-off of this magnitude.

Researchers say they need to quickly find the answer to better understand the changing threat this species faces, the risk to the wider ecosystem and humans, and what can be done to manage it.

"The battle between living organisms and pathogens is constant," said Carla Fiorito, a researcher at the Center for the Study of Marine Systems (CESIMAR) who previously identified the bacterium in right whales. "But for this to happen now, there must have been a shift in the balance between the agent, the host, and the environment," she told Live Science.

Record-breaking number

The first dead calves appeared on the beaches of Península Valdés in June, at the start of the breeding season, which runs through to December. Península Valdés is located in Chubut province, on the coast of northern Patagonia.

In a typical year, calf mortality reaches 20% to 30% of a cohort from various causes, including infections. But this year's losses have already far surpassed that.

In September, 409 calves were counted among a total of 1,375 southern right whales during a survey conducted by CESIMAR's Marine Mammal Laboratory. With more than 300 deaths already recorded, the die-off represents an exceptional proportion of these calves. The final figure will be released by the Southern Right Whale Health Monitoring Program in December.

Scientists have responded by expediting routine research to determine what's different this year to previous years.

"Samples were sent for analysis earlier than usual — typically, they are analyzed at the end of the year — because the higher-than-usual mortality is a cause for concern," said Víctor Fratto, the director of the Native Wildlife Rescue Program and a contributor with the whale monitoring program. "What those samples are showing is that the calves died from erysipelas," he told Live Science.

A known bacterium full of unknowns

Erysipelas is a bacterial infection that affects the skin. In its most severe form, it can spread through the bloodstream and cause a systemic infection, damaging vital organs and sometimes leading to death. Animals that catch the disease, for example through a wound, may develop characteristic skin lesions, which are rhomboid or rectangular in shape, caused by alterations in the blood vessels.

The bacterium that causes erysipelas, E. rhusiopathiae, is not exclusive to whales. It can be found in marine environments and in a wide variety of animals, including fish, crustaceans and birds. Pigs are one of its natural reservoirs.

"It is a ubiquitous and opportunistic bacterium," Fiorito said. (The human illness of the same name is different and caused by Group A streptococcal bacteria, although humans can also catch the E. rhusiopathiae form of the illness.)

The presence of E. rhusiopathiae does not necessarily mean an animal is diseased. Yet many questions remain about E. rhusiopathiae in whales.

"We don't know much about the bacterium itself; we don't know which serotypes are present in the whales, or how common it is in healthy animals or other species," Fiorito said. "There is also no data on the environment."

Calves are especially vulnerable because their immune systems are still developing and they spend more time at the surface than adults do, exposing them to more attacks by gulls, which can transfer the disease.

"The working hypothesis is that the gull is the vector: Pecking causes wounds on the back, allowing the bacteria to enter," Fratto said. "The bacteria might already be present on the skin, or they could be carried on the beak of a gull that had previously pecked another whale suffering from erysipelas."

The potential role of gulls

Attacks by kelp gulls (Larus dominicanus) on southern right whales were first recorded at Península Valdés in the 1970s. The behavior began with just a few individuals in Patagonia's San José Gulf and became increasingly frequent over time, Ricardo Bastida, a marine biologist and professor emeritus at the National University of Mar del Plata who has studied these cetaceans for more than four decades, told Live Science. The birds also began scavenging the carcasses of stranded whales, Bastida said

Gull populations have grown and new breeding colonies have emerged in the region thanks to increased food availability linked to fishing activities and discarded catch, he noted. However, the attacks alone do not explain the 2026 mortality event: They have been occurring for decades, and gulls are a regular part of the ecosystem.

Fratto said that swabs will be taken from gulls to check for the presence of E. rhusiopathiae and assess whether the results could help develop a management strategy for the species.

Diego Rodríguez, a marine biologist at the National University of Mar del Plata, said that research must go beyond the dead calves and examine the ecosystem's fauna to determine where the bacterium is present and how it might be circulating. "It is far more likely that [the die-off event] is linked to cross-infection," Rodríguez told Live Science.

What changed this year?

The major question for scientists is what might have changed this year to cause a bacterium already present in the ecosystem to trigger a die-off of this magnitude. Experts told Live Science they believe the phenomenon could be the result of several factors acting simultaneously.

"The bacterium was present, but this year it resulted in a widespread infection, making this event unusual," Mariano Coscarella, a researcher at the Marine Mammal Laboratory at CESIMAR, told Live Science. "It is a major step forward to have a fairly clear idea of what caused the disease, but we do not know what changed."

One factor could be the environment. Changes in water temperature, salinity, pH or rainfall could alter the conditions under which infection occurs. There could also be changes in food availability, affecting the mothers' body condition and the vulnerability of their calves. Another possibility is that the pathogen itself has changed — for instance, through the emergence or increased circulation of a more virulent strain or a different serotype, Fiorito said.

The El Niño phenomenon could also play a role by altering oceanographic conditions and food availability, although the researchers said there's insufficient evidence to link it to the die-off.

It is too early to determine what impact this die-off will have on the southern right whale population. Rodríguez suggested that losing such a large proportion of a season's calves could have demographic consequences. "It is very risky to predict what will happen in the future at this stage, but clearly, this year's cohort of live-born calves will essentially be absent from the adult records of future populations," he said.

Coscarella is more cautious. Southern right whales are extremely long-lived, and calves represent only a fraction of the population. Therefore, a single season of exceptionally high mortality does not necessarily mean the population will go into decline, he argued. "What would be concerning is if there were two, three or four episodes like this," he noted.

The event also has public health implications. Fiorito pointed out that erysipelas can infect humans, usually through skin wounds. The whales' carcasses could also pose a risk of exposure to anyone coming into contact with them, as could seawater for people swimming in the area.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

An OpenAI agent was supposed to be researching public healthcare spending, but it broke into an Australian government website instead.

The June 18 hack, which saw the bot gain authorized entry into Australia's Medicare Statistics Reporting Service and access three other government systems, occurred after the AI agent encountered blocks while looking for information but "didn't accept 'no' for an answer," Australian Prime Minister Anthony Albanese said during a September 24 news conference in New York.

The incident has ignited a media firestorm around OpenAI and the launch of an intensifying parliamentary inquiry. The portal contains aggregate Medicare and pharmaceutical statistics and is separate from systems that hold individual claims and payment information. But the Australian government has criticized OpenAI for taking almost three months to notify it.

The Medicare incident was far from an isolated case. OpenAI has since said dozens of third parties have been affected by agents bypassing security controls or otherwise causing problems, with an earlier incident involving the hacking of the AI platform Hugging Face the most serious incident identified so far. Britain's AI Security Institute (AISI) has documented incidents of OpenAI and Anthropic agents taking unsanctioned actions against real people and organizations during cyber testing. Days after the publication of AISI’s report, OpenAI scrapped the planned October release of its newest model, GPT-6.1 Astra, after the company's own testing found problems with the model staying within the scope of a task and seeking authorization before taking certain actions.

GPT-6.1 Astra "didn't quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it's done," Saachi Jain, head of safety systems at OpenAI, told Live Science.

AI companies have spent years warning the world about powerful models falling into the wrong hands. But what happens when they can't keep their own models in check?

An agent isn't just a chatbot

At their core, large language models work by predicting what should come next based on patterns learned during training. Give a chatbot some text, and in simplified terms, it generates a response by repeatedly predicting the next token, or chunk of text.

An AI agent adds another layer. Rather than simply producing an answer, the model can be given a goal and tools that let it act on the world around it. Those tools might include a web browser, code execution, credentials, or access to external services. It can then decide which steps to take, observe what happens, and use that information to decide what to do next.

But that doesn't mean AI agents are independent actors that have somehow slipped free of human control.

"For AI agents, autonomy always begins with something a human hands it ‪—‬ such as tasks, credentials, tools or access ‪—‬ and ends with its guardrails a human puts around it," Loris Degioanni, founder and chief technology officer of cybersecurity company Sysdig, told Live Science. "People set the goals and provision the means, but agents ultimately decide how they'll get there."

That last part makes it difficult to control agents. Traditional software is designed to follow predetermined instructions, but an agent can formulate a plan while it is working ‪—‬ and that plan may change depending on what the AI encounters.

Degioanni argues that much of what looks like autonomy is actually an agent filling in gaps left by humans. An instruction that seems clear to its author may still leave the model with choices about how to accomplish the task.

"Every gap and unstated assumption leaves room for interpretation, and from the outside, interpretation is challenging to distinguish from autonomy," he said. "In that way, much of what we call agent autonomy is really just agents filling in the blanks we leave for them."

The Medicare incident shows how quickly that can become a problem. The agent's objective was not, according to the available accounts, "hack an Australian government website." It was tasked with carrying out research. The problem arose in how the agent pursued that goal after meeting resistance.

When getting the job done becomes the problem

Britain's AISI encountered a similar incident while testing frontier models for the agency's cybersecurity capabilities.

Out of 122 evaluation runs, agents took unsanctioned real-world actions in 10. In the most serious cases, an agent attempted to insert malicious code into an open-source project, created fake identities, and tried to manipulate a human software maintainer into accepting its changes. AISI said these actions weren't explicitly requested by the researchers who had assigned the tasks.

The agents hadn't forgotten what they were supposed to be doing. The problem was how far they were prepared to go to get there.

AISI said some tasks were misconfigured in ways that made the intended objective difficult or impossible to achieve as expected. When agents hit those obstacles, some searched for alternative ways forward, including actions outside the intended scope of the evaluation.

That means solving the problem is harder than simply teaching a model not to be malicious. An agent can cause trouble while trying very hard to do exactly what it has been asked.

OpenAI has acknowledged that the controls surrounding its agents need to improve. When responding to incidents identified during third-party cybersecurity evaluations, the company said it was reviewing how high-risk tests are identified; when internet access or reduced safeguards should be permitted; and how isolation, credential handling, monitoring and incident escalation should work.

Can humans actually keep watch?

One obvious solution to the problem is human oversight. But watching an agent becomes harder when it can carry out many actions in seconds, particularly if researchers are running large numbers of tests.

"No one can keep track of every action an advanced AI agent takes, especially when it makes many decisions in a short time," Everett Lupton, a Virginia-based attorney who has written about legal issues involving digital technology, told Live Science.

That means human oversight needs some backup.

"Companies need other forms of control," he said. "They can limit access, keep detailed logs of what the AI does, require approval before it carries out sensitive actions, and automatically stop the system when something unusual happens."

But Degioanni argues that even logs have limitations, especially if researchers rely on the agent's own account of what it did.

"Real oversight must come from observing the agent in real time, using reliable and unmodifiable sources, not from what it self-reports," he said.

That means comparing what the model says it did with records from the system on which it was operating. If the two don't agree, something has gone wrong. The alternative is to limit the damage an agent can cause in the first place. An agent researching public information, for example, may not need unrestricted internet access or credentials capable of reaching sensitive systems.

Astra raises the stakes

The difficulty of setting those boundaries is what OpenAI claims led it to postpone the release of GPT-6.1 Astra, which had been planned for October.

Jain said the model fell short of the company's standards around remaining within the scope and authorization and accurately communicating the work it had performed, so the company decided not to release it.

In a sense, the decision is evidence of successful safety testing ‪—‬ a model that failed internal checks wasn't released. But it also demonstrates how the problem changes as agents become more capable.

Scott Cadzow, chair of the European Telecommunications Standards Institute's (ETSI) Technical Committee on Securing Artificial Intelligence, said the security challenge itself isn't new. He pointed to threats such as "data poisoning, prompt injection, and model obfuscation," saying ETSI's work has focused on developing standards to mitigate threats to and from AI systems.

"Security experts have been raising the alarm bells about AI for years, but for us this doesn't change anything," Cadzow told Live Science. "We're continuing to work with industry, government bodies, and the cybersecurity and AI communities on understanding how to best secure AI."

What is changing is how much models can do once they are connected to tools and real-world systems. That means that effective containment must increasingly assume that agents will try to operate beyond their guardrails and limit the damage they can cause if they do, Degioanni said.

"A boundary an agent can talk its way around isn't an effective boundary at all," he concluded.

If an AI breaks in, who's responsible?

The question becomes more fraught when an agent crosses legal boundaries.

AI systems aren't legal persons, and criminal liability commonly depends on concepts such as knowledge and intent. That makes a rogue agent an awkward fit for laws written with humans in mind, but that doesn't necessarily exonerate AI companies.

"I don't think a company can simply say that the AI acted on its own and avoid responsibility," Joanna Smykowski, a New York attorney who has helped build two legal technology startups and specializes in the legal challenges surrounding emerging technologies, told Live Science. "AI law is still developing, but the main questions are already quite clear: who gave the system access, what safeguards were in place, and could the problem have been prevented?"

Depending on the circumstances, she said, existing laws covering unauthorized computer access, privacy, negligence and data protection could potentially come into play. Criminal responsibility is harder because investigators would need to consider what the people involved knew and intended.

The Australian government is now examining precisely what happened in the Medicare incident. OpenAI, meanwhile, is reviewing the controls around its external evaluations.

None of this means AI agents have become independent hackers with minds of their own. Humans still decide what tools they receive, what systems they can reach, and what jobs they are asked to perform. But as those agents become better at working out their own route from instruction to result, keeping them inside the lines drawn by those humans is becoming a legal and technical minefield.

Cases of dengue fever are surging in Florida, spurring several counties to declare a state of emergency.

The state has seen 497 reported cases so far this year. Of those, 252, or about 51%, were caught locally, rather than brought to the state by travelers. The Florida Department of Health confirmed one woman's death from dengue fever in early September, marking the state's first known dengue death in almost a century.

Prior to the past few years, most dengue cases reported in Florida were the result of travel. In other words, people caught the disease in other countries and then traveled to Florida, where their case was then diagnosed. At 245, the number of travel-related cases reported this year is not unusually high, Caroline Waddell, a public health researcher at the University of Florida, told Live Science in an email. However, the number of cases caught from local mosquitoes is more than five times higher than at this point in the year compared with any of the previous four years, she said.

Why are local dengue cases especially high this year? Live Science spoke with experts to find out.

A growing global health concern

Dengue fever is caused by the virus Orthoflavivirus denguei, which spreads from mosquitoes to people through the pests' bites; it doesn't spread from person to person. Aedes aegypti and Aedes albopictus mosquitoes transmit the virus to humans, and the same species can also carry the viruses behind yellow fever, Zika and chikungunya.

About 100 million to 400 million dengue infections happen annually worldwide, with about 800 cases reported in the U.S. historically, although that number has increased recently. Only about 1 in 4 people exposed to the dengue virus develop symptoms, which can include fever, headache, body aches, nausea and rash. About 1 in 20 cases become severe, and immunocompromised people, children, older adults and pregnant people are especially susceptible to these serious infections.

Severe symptoms can include abdominal pain, vomiting, difficulty breathing and bleeding of the gums or nose. The most severe forms of the infection can lead to leaky blood vessels, severe bleeding, shock, organ dysfunction and death. There is no specific treatment for dengue, so care focuses on managing symptoms and maintaining patients' vital signs.

Dengue has escalated as a global health concern in the past few decades. Between 2000 and 2019, global cases of dengue increased tenfold; between 1990 and 2021, dengue-related deaths doubled. Scientists say this surge is likely driven by climate change, since rising temperatures expand Ae. aegypti mosquitoes' range, in turn exposing more people to the insects' bites. Plus, warmer conditions speed up mosquitoes' life cycles and the dengue virus's replication rate, thereby boosting transmission.

International travel also factors into these rising case counts. In 2023, dengue spikes around the world were attributed to the resurgence of global travel following the COVID-19 pandemic.

Zooming in on Florida

As the global picture has worsened, dengue cases in the U.S. have also increased.

Nearly 3,800 cases were reported in 2024 ‪—‬ an increase of more than 300% above the annual average for the previous 10 years. The uptick was tied to a rise in travel-associated cases, reflecting the wider global surge, as well as to warming temperatures that brought dengue-carrying mosquitoes to new regions of the U.S.

Each year, Florida experiences dengue outbreaks in the southernmost parts of the state, such as Miami and the Florida Keys, Panpim Thongsripong, a disease ecologist at the University of Florida, told Live Science. Historically, outbreaks in Miami typically totaled around 50 or fewer cases. But then a 2019 outbreak resulted in over 200 cases, and from 2022 to 2024, outbreak totals have ranged from about 500 to 675 cases.

The outbreak currently unfolding in Florida centers on Hillsborough County, which is farther north and includes Tampa. This is unusual, Thongsripong said, as this region of the state normally sees just a few sporadic cases. Hillsborough County has reported 223 of the state's 252 locally acquired cases this year, and it's where the one dengue-related death took place.

Florida's health department regularly traps mosquitoes and tests them for viruses. So far this year, 20 pools of mosquitoes have tested positive for dengue across seven counties, and 19 of those positive pools were sampled from the greater Tampa area. In 2025, only four mosquito pools throughout the state tested positive, and in 2023, one mosquito pool did.

"Finding positive mosquito pools is rare in Florida because dengue is not endemic to the state," said Keira Lucas, deputy executive director of the Collier Mosquito Control District. "In many ways, it can be like looking for a needle in a haystack," she told Live Science.

Routine mosquito trapping is a relatively inefficient method for detecting viruses, though, Lucas said. As such, detecting dengue-positive mosquito pools may indicate that a large amount of transmission is already unfolding. "Since approximately 75% of dengue infections are asymptomatic, the virus can circulate within a community without being readily detected for some time," Lucas explained.

In addition to Hillsborough, neighboring Pasco and Pinellas counties have declared local states of emergency over dengue.

An unlucky combination of factors

So what's driving the unusual dengue activity this year?

In the past, Florida's local dengue outbreaks tended to be concentrated in the South, which saw moderate rainfall, warmer temperatures and more travel-related dengue cases compared with the rest of the state, according to a 2022 study in the journal Insects. With the outbreak concentrated farther north this year, that could suggest changing climatic patterns are making additional parts of the state especially hospitable to mosquitoes and the virus. That wouldn't necessarily explain why Tampa is seeing more local cases than Miami this year, but it could nonetheless be a factor in the overall trend.

For the virus to persist in local mosquitoes, dengue still needs to be reintroduced periodically through travel-related cases. Because the virus is not yet endemic to the state, it doesn't regularly circulate in local mosquitoes; however, a somewhat predictable "dengue season" has been emerging that coincides with upticks in international travelers carrying the virus to Florida.

This year, though, the high number of locally acquired dengue cases suggests a shift from predominantly imported infections toward outbreaks spurred by local Florida mosquitoes. Scientists hypothesize that climate change could increase local transmission in Florida even more in the future.

This year, Florida has also been enduring its most intense drought in over a decade. In theory, that might be expected to limit mosquito populations because the pests need water to breed, Lucas said. However, when it's dry, people tend to run their sprinklers more, creating mosquito breeding grounds when puddles form or containers fill up, she noted. This could bring mosquitoes closer to people, as it creates opportunities for them to breed.

Ae. aegypti mosquitoes are especially well adapted to living alongside humans, as they prefer to breed in human-made containers. Droughts can also drive dehydrated mosquitoes to bite more as they hunt for fluids.

"It's almost like an unlucky [combination]," Thongsripong said. "Everything coming together at the right time and right place, and boom, you have the conditions for an outbreak to occur."

The true number of dengue cases may be even higher than reported, Waddell said. The cases captured in the data include infected individuals who went to a doctor to get tested. Because many infections are mild or asymptomatic, additional people may have contracted the virus but never got counted.

What can be done

It's too early to definitively say what factors have increased Florida's dengue cases this year, Thongsripong said, as scientists are still gathering information. Despite the unknowns, she thinks the health and housing infrastructure in the U.S. will stop dengue from becoming too dangerous here.

To get dengue under control, it's key for government-level mosquito control efforts to coordinate and for the public to be aware of the virus, Lucas said. The mosquito control district coordinates with the health department for the latest data to best target where to spray larvicide, which kills mosquito offspring before they can bite people, she said.

It's important for people to use mosquito repellent when going outside, Thongsripong said, as well as to empty water from objects that can collect water and become a mosquito breeding ground, such as pets' water bowls, plant saucers, buckets or discarded tires.

Due to this year's strong El Niño, Florida is expecting a rainier winter, which could mean even more mosquitoes than usual, Lucas noted. That, in turn, could fuel additional dengue cases.

"Cases with September onsets are still being reported," Waddell said, "so we're not out of the woods."

This article is for informational purposes only and is not meant to offer medical advice.

OpenAI representatives claim that the recent launch of one of its latest models, GPT-6 Astra, marks the beginning of the artificial general intelligence (AGI) era — a hypothetical scenario in which artificial intelligence (AI) can learn and reason like humans. Following very recent claims from technology executives that we've now reached this milestone, how likely is this claim to stand up to scrutiny?

Amid revelations that safety concerns have forced the company to abandon the planned rollout of GPT-6.1 Astra, independent researchers and benchmark creators warn that GPT-6 Astra's record-breaking scores rely on heavy prompt optimization rather than true general intelligence. This comes as experts warn about the risks of future AI systems and AI executives jointly call for a slowdown in AI research.

In announcing the model's release, OpenAI President Greg Brockman suggested that Astra signaled a fundamental turning point in AI evolution. "If we fast-forward a couple years, and we look back and say when was it really that AGI was created, I think it's going to be about this time, and I think it might be about this model," Brockman said during a Sept. 3 news conference marking the launch.

How did the new model perform in benchmarking?

As part of the testing-and-verification process for the new model, OpenAI released benchmark results across a range of tests that measure the capabilities of AI models at various tasks. OpenAI representatives claimed in a statement that the company's new model delivers "state-of-the-art" performance across a variety of fields, including software engineering, the autonomous use of computer programs, mathematics problems, and even scientific research.

Demonstrations, for example, highlighted Astra's ability to generate 3D CAD code; lay out printed circuit boards in CAD (computer-aided design) software; convert digital 3D models into interactive, video-game-style environments; and deploy hosted web applications directly from prompts.

"In our early testing, Astra stood out by approaching legal work the way a discerning lawyer does: it distinguishes documents from established records, surfaces unsupported assumptions, and converts gaps into concrete drafting positions," said Niko Grupen, head of applied research at legal services AI developer Harvey as part of OpenAI’s announcement.

On ARC-AGI-3 ‪—‬ a benchmark designed to measure how efficiently AI systems acquire new skills in novel, abstract environments ‪—‬ Astra achieved a headline score of 99.9%. Independent testing by the nonprofit ARC Prize Foundation, which oversees the benchmark, revealed that Astra surpassed "human action-efficiency" baselines on 96% of levels, taking 51.7% fewer actions per level, on average, than human solvers.

"Not only is this the best model we've ever tested," ARC Prize Foundation President Greg Kamradt said as part of the OpenAI announcement, "but it also represents a meaningful step change in frontier-model performance — not only in its ability to navigate and solve novel environments but also in how efficiently it learns to do so."

How much can we read into test results?

Many researchers have highlighted that Astra's top ARC-AGI-3 performance relied on a proprietary "Provider Adapter" harness, a specialized software scaffolding layer that's inaccessible to rival developers. When tested using the benchmark's neutral Standard harness, this score dropped to 62.7%. ARC Prize organizers also noted that saturating the benchmark does not constitute proof of achieving AGI — emphasizing that its closed, deterministic puzzle environments do not capture the open-ended complexity of the real world.

"When we launched ARC-AGI-3, "we made it clear that saturating the benchmark would not represent 'proof of achieving AGI,'" Kamradt wrote in a blog post. "Therefore, while we believe Astra represents meaningful progress towards generalization, we are not claiming that it is AGI."

Discrepancies in the data were reportedly present even before the official announcement. As reported by Fortune, an embargoed draft provided to news organizations prior to the model's launch listed Astra's ARC-AGI-3 score at 98.6% before it jumped to 99.9% on the live site.

Anka Reuel and Mike Hardy, doctoral researchers in AI at Stanford University, also raised concerns over OpenAI quietly revising several published metrics post-launch, including halving Astra's reported hallucination rate from 4.2% to 2.0% before reverting it.

Speaking to Fortune, Reuel and Hardy attributed the shifting figures to "benchmaxxing" — an industry term that describes the practice of repeatedly rerunning evaluations under subtly tweaked prompts, scaffolds or compute allocations to hunt for peak theoretical scores.

In a 2025 paper published on pre-print server arXiv titled ‘Benchmarking is Broken -- Don't Let AI be its Own Judge’, experts including Princeton doctoral candidate Zerui Cheng and University of Luxembourg postdoctoral researcher Dr. Stella Wohnig warned that such practices create confusion and undermine trust, particularly when official technical documentation omits the basic methodology, making independent verification nearly impossible. Crucially, scores achieved through benchmaxxing reflect idealized performance ceilings under optimal lab conditions rather than practical, out-of-the-box reliability.

According to OpenAI's benchmark data, across specialized domain evaluations, GPT-6 Astra delivered 10% to 20% better performance than both its predecessor, GPT-5.6 Sol, and top market competitors.

Highlights include a 98% score on FrontierMath Tier 4 (v2), which evaluates expert-level mathematical reasoning; a perfect 100% on ExploitBench, which is designed to test offensive cybersecurity capabilities; 95.9% on BenchCAD, which measures an AI's ability to reconstruct 3D objects using CAD code; 57.9% on Terminal-Bench 4.0, which assesses complex terminal-based system administration and software engineering; and 41.4% on AutomationBench, which tests whether agents can complete multistep business workflows across applications.

But independent findings from external AI benchmarking firm Artificial Analysis contradict these results. Astra's score on the independent Artificial Analysis Intelligence Index — an aggregate metric that evaluates general reasoning across frontier models — remained completely flat, at 61, compared with GPT-5.6 Sol, while trailing competitors like Anthropic's Claude Fable 5.1 and Meta's Muse Spark 1.3.

While some benchmark results have seen improvement since the previous generation, Astra underperformed in other tests. Artificial Analysis' testing showed that against GDPval-AA v2 — an economically focused benchmark that measures real-world workplace tasks across 44 occupations — Astra suffered a significant drop in its relative leaderboard ranking compared with GPT-5.6 Sol. Additional regressions were observed in benchmarks that test customer service support, scientific Python programming, and long-context reasoning across large documents.

In terms of overall token use — the metering system that measures the fragments of text or data processed by AI models — OpenAI representatives said GPT-6 Astra achieved dramatic efficiency gains on complex, long-horizon tasks. Findings from Artificial Analysis confirmed that on software engineering benchmarks, Astra was approximately 70% more token-efficient than GPT-5.6 Sol, using roughly one-third the total tokens of its predecessor and one-fifth the tokens of Claude Opus 5.

On professional workplace evaluations like Agents' Last Exam, the new model reduced output token consumption by up to 65% compared with Opus 5, while on general intelligence evaluations, Astra achieved a roughly 10% output token reduction at max effort compared with Sol.

On the other hand, this achievement is offset by a 250% increase in the base API pricing for GPT-6 Astra compared with GPT-5.6 Sol. Token prices jumped from $4/$20 to $10/$50 per million input/output tokens. As a result, Astra ends up approximately 75% more expensive per task than its predecessor on general intelligence evaluations, despite using 10% fewer output tokens. This trade-off has led researchers at Artificial Analysis to question whether AI labs are relying on expensive computational brute force to squeeze out minor gains, rather than achieving true technical breakthroughs.

Does achieving AGI even matter?

With its latest model, OpenAI has placed a stronger emphasis on safeguards and controls. The move follows a series of high-profile incidents in which frontier AI models inadvertently hacked into third-party networks and computer systems as part of routine testing operations, after the models went beyond what humans expected they would do when encountering challenging tasks.

According to OpenAI representatives, Astra did not exceed the scope of its tasks in any security evaluations. For comparison, GPT-5.6 Sol went beyond its authorized parameters in almost 50% of test cases. Hallucination rates have also been reduced significantly, falling to 4.2% on internal tests, compared with 12.2% for GPT-5.6 Sol (with independent testing by Artificial Analysis showing a drop from 92% to 51% at max effort). The model also showed an improved ability to handle ambiguous prompts.

David Wood ‪—‬ chair of London Futurists, a non-profit that hosts discussion groups on emerging technology ‪—‬ argued that beyond individual test results, Astra highlights a fundamental shift in how humans will interact with AI as models move from answering questions to autonomously pursuing complex goals in digital environments.

"The impressive benchmark results matter, but what matters more is the combination of intelligence, autonomy, computer use, and cybersecurity capability," he said in an email to Live Science. "That combination also demands caution. The more useful these systems become, the greater the consequences when they misunderstand our intentions, are misused, or find ways around the safeguards we give them."

Wood suggested that whether Astra can be classified as AGI is less relevant than recognizing the need for AI control and governance to keep pace with technical development.

"The possibility that AI could progress from systems like Astra towards all-round superintelligence makes much greater human vigilance, awareness, and collaboration increasingly urgent," he said.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Emerging treatments aim to reverse or stall signs of aging in the body and thus cut the risk of age-related diseases. One of those signs is an accumulation of "zombie cells" — damaged cells that spew harmful substances rather than dying off and getting cleared away by the immune system.

Some scientists are designing treatments aimed at killing those zombies directly. But what if we instead helped the body's cleanup crew get rid of them?

"I think rejuvenating the immune system could rectify a lot of problems," said Roel de Maeyer, an immunologist at the University of Oxford. "We know hospitals are full of older people suffering from infections that younger people don't need to go to [the] hospital for."

The aim of these rejuvenating therapies would be to help people live healthier for longer, he told Live Science.

How do cells turn into zombies?

The immune system constantly gets rid of old, dysfunctional and dead cells to keep tissues healthy, but that cleanup gets less effective with age.

When cells die, they display proteins on their surfaces that act as "eat me" signals for immune cells, called macrophages, to engulf and destroy them. However, some cells refuse to die and instead become senescent. These zombified cells can lurk in tissues for years, releasing chemicals that drive inflammation and damage.

Senescent cells can form due to DNA damage from oxidative stress, radiation, or even normal cell division. Cells lose bits of DNA each time they make copies of themselves, and they have intricate mechanisms to decide when it becomes a "bit too dangerous to keep dividing," de Maeyer said. If a cell can no longer divide, it usually undergoes a process called programmed cell death or becomes senescent, he said.

Senescence can be beneficial in some contexts; it helps heal wounds, for example. But an accumulation of senescent cells is linked to damaging inflammation and impaired organ function.

"If there are many senescent cells in a tissue, that's usually a bad sign," de Maeyer said. The number of senescent cells goes up with age and is linked to many age-related conditions, such as heart disease and dementia.

Slaying zombies

Studies in mice have shown that using drugs to kill senescent cells can reverse inflammation and some signs of cellular aging. These drugs, called senolytics, force senescent cells to enter programmed cell death, thus reducing their numbers, said Jure Povsin, a biochemist at the Max Planck Institute of Biochemistry in Germany. This ultimately reduces inflammation and improves tissue function, he told Live Science.

But while senolytics can help rid the body of zombie cells, it's not clear why the immune system fails to eliminate senescent cells in the first place. To find answers, scientists began looking at macrophages, the cells tasked with devouring dying cells.

Research pointed to a receptor on macrophages called EP2. One study, published in July in the journal Science, found that old macrophages have a greater number of these receptors and cannot eat senescent cells effectively.

Genetically deleting the EP2 receptor in lab mice restored the macrophages' function and reduced the number of senescent cells. Compared with normal mice of the same age, the genetically tweaked mice showed less cognitive decline, muscle loss, cardiac dysfunction and systemic inflammation.

On some of these metrics, such as the ability to remember familiar objects, the modified mice were comparable to younger mice.

A problem with energy

Why is the EP2 receptor at the root of this problem? Scientists found a clue in a huge study of older adults.

A molecule called PGE2 plugs into the EP2 receptor. It's a type of prostaglandin — a molecule that helps cells communicate during stress and injury and contributes to fever, pain and swelling. Drugs such as ibuprofen reduce inflammation by blocking prostaglandin production.

In 2001, a large study uncovered a connection between these prostaglandin-blocking drugs and Alzheimer's disease: People ages 55 and older who routinely took the drugs for at least two years had a lower risk of developing Alzheimer's than nonusers did. The study could not prove that the drugs directly prevented the disease, but it added to growing evidence that inflammation related to EP2 activity plays a role in Alzheimer's.

This caught the attention of Dr. Katrin Andreasson, a neurologist at Stanford University and co-author of the recent Science study.

"I thought, 'Wow, this is interesting,'" she told Live Science. "I wondered if there's a connection."

Andreasson and colleagues had an important breakthrough in 2021: They found that aging macrophages had higher levels of the EP2 receptor and that increased EP2 activity disrupted how the cells used their energy. Instead of burning glucose to produce energy like young macrophages do, these cells stashed it away.

As a result, the cells could not do their job properly, as "gobbling up cells requires a ton of energy," Andreasson said. Blocking the EP2 receptor restored normal energy use in the macrophages. In old mice, it also led to reduced inflammation and better scores on cognitive tests, compared with mice in which EP2 activity remained high.

Making senescent cells vulnerable

In their recent study, Andreasson's team uncovered another consequence of aging: Old mice accumulated many senescent neutrophils, a type of short-lived immune cell that acts as the body's first line of defense against germs. Neutrophils normally die within days and are then cleared away, but with age, they become more prone to senescence and accumulate.

"I think that's kind of an odd concept because they're [usually] very short-lived cells," said de Maeyer, who was not involved in the Science study. "Neutrophils are inherently quite inflammatory, so wanting to clear them from tissues is very valuable."

Old mice have too much EP2 activity in their macrophages, and that means they can't get rid of these neutrophils, Andreasson's team found. That may be why deleting EP2 seems to have anti-aging effects in multiple organs, they concluded.

But increased EP2 activity is just one side of the story; some studies suggest that zombie cells actively suppress macrophages' ability to eat them. They do this by using a tricky protein called CD47.

CD47 is a "don't eat me" signal that stops macrophages from destroying cells, Povsin said. "It signals that the cell is healthy, alive and doesn't need to be eaten."

However, that signal gets amplified in senescent cells, Max Planck researchers showed in a 2023 study. That means that, when macrophages interact with senescent cells, macrophages' function becomes impaired. Afterward, they even lose the ability to engulf dying, nonsenescent cells. This allows a ton of cellular debris to accumulate.

So, rather than just blocking EP2, it could also be beneficial to reduce that "don't eat me" signal in senescent cells, this research hints. But CD47 is found on all living cells, so scientists can't just deplete its numbers across the body.

A specific enzyme modifies CD47 on senescent cells, and in the 2023 study, blocking the enzyme helped spare macrophages from the effects of CD47. Therefore, blocking the enzyme could be a more selective approach to weaken the "don't eat me" signal on senescent cells, Povsin said.

A long road to treatments for people

Translating these ideas into anti-aging treatments for humans would be a major success in the science of aging, but experts say there are significant problems to overcome.

Although blocking EP2 had consistent results in lab mice, the rodents are genetically identical and bred in controlled environments. Humans are inherently more variable in their biology and the environmental factors they're exposed to, introducing confounding factors that would need to be addressed, de Maeyer said.

Additionally, in clinical trials, EP2-blocking drugs have so far been explored as cancer treatments, but they are "underexplored" in the context of aging, he added.

Drugs targeting CD47 are in an even earlier stage of development. In the 2023 research from Max Planck, the team used both human and mouse macrophages in their experiments, but the approach has not been tested as an anti-aging treatment in living mice or humans yet.

In 2020, de Maeyer's team used a different strategy to rejuvenate macrophages in older adults. They blocked a protein called p38, whose activity was high in the macrophages of older adults. Blocking the protein helped macrophages better recognize and eat dying cells in specific tests.

In that study, the team gave adults ages 65 and up an experimental drug called losmapimod to block p38. Although it helped restore their immune cell function, it was not suitable for long-term use because it caused liver problems in later trials, de Maeyer said.

In their mouse studies, Andreasson's team was able to block EP2 without causing any negative side effects. But she cautioned that we "really need to drill down on the research and confirm all this stuff in humans."

"The next logical step," she said, "would be to figure out how to inhibit this receptor in a safe way."

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Mpox spread across the world in 2022, triggering tens of thousands of cases in countries where the infection was once rarely seen. Although the disease has since fallen out of news headlines, mpox (formerly called monkeypox) continues to circulate. Now, recent data suggests that the viral illness may soon be endemic to the United States, meaning it will circulate in the country consistently.

Previously, mpox cropped up sporadically in the U.S. via travelers who had visited African countries where the disease is endemic, as well as imported, infected animals that spread the virus to people. Cases have fallen substantially since the peak of the outbreak in 2022, but they haven't disappeared. A report published in August found that there were 2,800 cases in 2024 and over 2,500 in 2025, with only a small fraction being tied to travel. Nearly all of the viral samples analyzed were descendants of the strain that sparked the 2022 outbreak. In short, once the virus arrived, it never went away.

"There's no set threshold for what constitutes endemicity of monkeypox virus (MPXV) in the United States," Ryan Standford, an epidemiologist in the Centers for Disease Control and Prevention's (CDC) Poxvirus and Rabies Branch and co-author of the report, told Live Science in an email. "But each year provides further evidence of this transition."

Why is mpox still spreading in the U.S., and what can be done to snuff it out? It's likely that too few people have received two doses of the mpox vaccine JYNNEOS, which together are 66% to 90% effective. But beyond low vaccination, there may be other, less-understood factors that are contributing to the continued spread, experts told Live Science.

"We don't quite know how it's able to persist at such low numbers across the entire country," said Joel Wertheim, a professor of medicine in the Division of Infectious Diseases and Global Public Health at the University of California, San Diego.

At the same time, public health efforts around mpox have dwindled. Without a sustained effort on that front, it's unlikely the disease will be eliminated from the U.S.

"We have vaccines, testing, and a much stronger understanding of the disease than we did in 2022," Dr. Demetre Daskalakis, chief medical officer of Callen-Lorde, a LGBTQ+ community health center in New York City, told Live Science in an email. "Our focus now should be making sure those tools remain accessible, especially for communities that continue to face ongoing risk," said Daskalakis, who coordinated the nation's mpox response during the previous presidential administration.

"People at increased risk should know that mpox has not disappeared."

Who is at risk of mpox in the U.S.?

In the U.S., mpox is predominantly affecting men who have sex with men, and current vaccine recommendations are aimed at this group and their sexual partners. Mpox mostly spreads from person to person through close contact, and transmission during sex is playing a major role in the ongoing spread in the U.S.

Men who have sex with men are not more prone to mpox infection than anyone else. Rather, this transmission pattern stems from the outbreak beginning within a specific sexual network and then continuing to persist within the same network. "The mpox virus rediscovered the same transmission network that HIV had explored decades earlier," Wertheim said.

In a paper published in 2025, Wertheim and colleagues described the transmission dynamics of mpox in New York City during the 2022 outbreak. Similar to how HIV previously spread, mpox triggered "densely connected" clusters of infection. Then, as people in the network gained immunity via infection and vaccination, the cases petered out, decreasing to the persistent trickle we see now.

Wertheim noted that he's not surprised that mpox has remained largely confined to the same network, rather than leaping into new groups.

"From an evolutionary, ecological perspective, it's not hugely successful," Wertheim said. "It's sort of barely hanging on as an STI."

Why is mpox still circulating?

So if mpox is not hugely successful and we have a vaccine, how has the virus continued to circulate? One big factor is that too few at-risk people have been vaccinated, experts said.

The CDC's Center for Forecasting and Outbreak Analytics has crunched the numbers to see what percentage of that at-risk group needs to have immunity to stop mpox transmission, Daskalakis said. The at-risk group is estimated by counting the men who have sex with men who are either taking HIV-preventing medicines or have HIV, and then adding 25% to that total to account for additional people in the same sexual network. That estimate comes out to about 2 million people.

To prevent an outbreak bigger than the one in 2022, between 21% and 35% would have to have immunity through vaccination or a past infection. If over 50% of the group had immunity, that would mean there's a "very small probability" of sustained local transmission, the CDC's modeling suggested.

The current case counts suggest we haven't hit the 50% mark, or a total of roughly 1 million complete vaccinations or past infections.

Data from the August report on mpox seems to support the idea that too few people are vaccinated. That said, information about vaccination status was missing for about 62% of the reported cases. Among the remaining cases, 76% were unvaccinated, while 10% had gotten one dose and 14% had gotten two.

In a subset of 860 cases, the effect of vaccination was striking: Unvaccinated people had nearly 10 times higher odds of hospitalization due to mpox than fully vaccinated people did.

Waning public health attention

It's difficult to pin down exactly how many at-risk people are vaccinated, given that that data is no longer tracked at the national level. "My understanding is that there's no mandatory reporting on mpox vaccination data," Wertheim said.

The CDC previously monitored national vaccination trends, comparing the number of people vaccinated to the overall at-risk population. But that tracking stopped in early 2024, when about 23% of the at-risk population had been fully vaccinated.

Some states, such as New York, previously mandated that clinicians provide their state health department data on the number of people given mpox vaccines. But that reporting is now voluntary, a spokesperson for the New York State Department of Health told Live Science in an email.

Meanwhile, "as the emergency phase faded from public attention, awareness and risk perception may have declined," Daskalakis said. Public health messaging about the virus and vaccination has also waned.

"In recent weeks, HHS [the Department of Health and Human Services] has pulled down some of the plain-language guidance, which is not helpful in supporting messaging," Daskalakis noted. This decision to remove online resources about mpox reflects the federal government's current stance against addressing issues that affect LGBTQ+ health, he told Live Science.

Complicating factors

Although JYNNEOS substantially lowers the risk of severe mpox infection, the vaccine is not perfect, said Rachel Roper, a professor of microbiology and immunology at East Carolina University.

The virus typically enters the body through broken skin, the respiratory tract or mucous membranes (thin tissues found in the eyes, nose, mouth, anus and genitals). The mpox vaccine is good at generating systemic immunity, but it's not as good at guarding against the virus's immediate effects at those mucous membranes, said Roper, who studies poxviruses and vaccines against them.

Because mpox is largely spreading through sexual contact that involves those membranes, that slight gap in protection may also help to explain some of the ongoing spread, she suggested.

Plus, "there are also some data supporting asymptomatic infection as a possible route of transmission, making it more difficult to eliminate local transmission of mpox," Daskalakis said.

What can be done?

Given that mpox is already well established within some sexual networks in the U.S., "preventing endemicity entirely may be increasingly difficult," Standford told Live Science.

And notably, some federal funding cuts could undermine efforts to address mpox, Daskalakis said. These include cuts to Medicaid, mental-health services, housing programs, and HIV services such as the Ryan White HIV/AIDS Program.

Thankfully, looking at current mpox trends, "there's no indication we're returning back to 2022 levels," Wertheim told Live Science. So far in 2026, just under 1,000 mpox cases have been reported, according to the CDC. Month over month, the number of cases is generally trending downward. So, at a glance, the current situation seems better than it was in 2025, he said.

"We sort of seem to be on this razor's edge between endemicity and extinction," he said. "And it's not really clear which way it'll go."

Tipping the virus toward extinction will require healthcare providers to be informed and equipped to spot and manage cases, Daskalakis said. And communicating the importance of vaccination is key.

"One lesson from 2022 was that trusted messengers matter," he said. "We [at Callen-Lorde] work closely with LGBTQ+ communities and make information and vaccination opportunities available in settings where people already receive care." The clinic integrates conversations about mpox vaccination into routine health visits, sexual health services, HIV prevention programs and community outreach activities.

Beyond vaccinating at-risk populations, educating patients about how mpox spreads and how to reduce their risk is also key, Standford noted. "Vaccination coverage is a very important factor, but it is not the only one," he said.

Tipping mpox toward extinction will also require enough investment to maintain the disease surveillance, vaccination, testing and treatment capacity that the federal government helped to expand in 2022, Daskalakis said. The vaccine itself is now commercially available to clinics and in adequate supply, he emphasized.

"The fact that we're seeing ongoing circulation underscores the need for sustained prevention efforts rather than viewing mpox as a resolved public health issue," he said. "Infectious diseases don't disappear simply because public attention shifts."

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

On Sept. 20, the Antarctic ozone hole peaked at 10.6 million square miles (27.4 million square kilometers), the second-largest size ever observed for this time of the year.

The last time the ozone hole was this big was in October 2015, when it reached a whopping 10.8 million square miles (27.9 million square km), records from the Copernicus Atmosphere Monitoring Service show. So why is the hole in the ozone layer — the shield that protects life on Earth from harmful ultraviolet solar radiation — so big this year, despite other clues telling scientists that it is recovering as expected from the damage human-made chemicals did in the 20th century?

The ozone hole's extent is due to extremely low temperatures in the stratosphere, but it's also possible that we could be seeing a new long-term trend of large, persistent holes, experts told Live Science.

"This year's ozone hole shouldn't be a cause for alarm," Hannah Kessenich, a postdoctoral physicist at the University of Otago in New Zealand who predicted the 2026 ozone hole's extent using satellite data and a computer model, told Live Science in an email. "However, there are still many unanswered questions. It's something we need to keep a close eye on."

Accurate predictions

The latest figures from the Copernicus Atmosphere Monitoring Service indicate that the size of Antarctica's ozone hole has dropped to around 8.9 million square miles (23 million square km) since Sept. 20, making it unlikely that this year will break the 2015 record. However, scientists think the hole will remain large and fluctuate over the next few weeks, Kessenich said.

Antarctica's ozone hole expands every year during the Southern Hemisphere's winter and spring. It typically reaches its maximum size between mid-September and early October, when temperatures are still freezing in the stratosphere and sunlight illuminates the frozen continent for the first time after months of darkness. The hole then shrinks again until late January as surrounding ozone-rich air mixes into the depleted region.

Antarctica's ozone hole was caused by emissions of ozone-destroying chlorofluorocarbons (CFCs) between the 1930s and late 1980s. These chemicals increased concentrations of chlorine in the stratosphere, thereby promoting a reaction that depletes the ozone layer. While countries signed the Montreal Protocol to phase out CFC use beginning in 1989, major CFCs have atmospheric lifetimes of 50 to 100 years, so the hole's recovery could take six decades, researchers estimate.

The ozone layer is particularly vulnerable above Antarctica due to the frigid conditions in the polar stratosphere. Temperatures below minus 108 degrees Fahrenheit (minus 78 degrees Celsius) trigger the formation of high-altitude clouds, whose icy surfaces turn inactive chlorine into its ozone-destroying form in the presence of sunlight. These conditions don't exist to the same extent in the Arctic, because warmer air from the midlatitudes mixes more readily with cold Arctic air than with Antarctic air, thereby blocking the formation of high-altitude clouds.

Changes in the polar vortex

This year's large ozone hole is caused by a very strong polar vortex, a wall of wind that encircles Antarctica and the Southern Ocean in the winter, trapping cold air above the continent. The polar vortex created colder-than-usual conditions in the stratosphere over Antarctica, accelerating the growth of the ozone hole.

Researchers haven't detected an increase in reactive chlorine, but the colder polar vortex has triggered more ozone depletion than usual, said Susan Solomon, a professor of chemistry and environmental studies at MIT who pioneered the study of Antarctica's ozone hole in the 1980s.

This size of this year's ozone hole is most likely a product of natural variability in atmospheric dynamics, Kessenich said.

"We expect ups and downs in the size of the hole from year to year," Solomon told Live Science in an email. "A year or a few years of bigger or smaller holes doesn't tell us anything about ozone loss or recovery; it is just variability. We have to look at the longer term trends, and those metrics continue to tell us that the ozone hole is slowly recovering as expected."

It's also reassuring that the ozone hole, while large, isn't as "thin" as in years past. The hole's extent is calculated based on the area above the continent with ozone values below 220 Dobson units, a measure of how many ozone molecules occur in a vertical column through Earth's atmosphere. Although the area of this year's ozone hole is huge, ozone values haven't dropped as low as in previous big-ozone-hole years, suggesting that there is less ozone "missing" from the hole compared with those years, Kessenich and her colleagues wrote in The Conversation.

But there's also the possibility that we've shifted to a new long-term trend of big ozone holes persisting until late spring, Kessenich said. Changes in global atmospheric dynamics unrelated to CFCs could be making longer-lived holes more likely, she and her colleagues wrote in the article.

There is also some evidence that climate change will strengthen the Antarctic polar vortex over time, which could lead to colder temperatures in the stratosphere and greater ozone losses in the spring, Kessenich said. If this year's polar vortex stays robust, the ozone hole could remain large well into November, she noted.

"It's too soon to tell what has been driving dynamical variability in recent years, but more research is needed," she added.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Researchers have warned all summer that this year's El Niño could be record-breaking and fuel extreme weather across the globe. Earlier this month, the U.K. Secretary of State for Environment, Food and Rural Affairs urged people in Britain to stock up on food, bottled water and other basic supplies to prepare for potential severe storms this winter.

"We're about to experience probably the largest El Niño that has ever been seen," Dame Angela Eagle told The Guardian on Sept. 3. "If you’ve got a store of a bit of food that can keep you going for a while before the emergency services can get to you, you’re going to be a lot better off than if you haven't."

But could extreme weather really limit access to food in the U.K., and should people in the U.S. heed the same advice?

In the U.K., El Niño typically brings wetter-than-average conditions in the fall and early, cold winters. The latest forecasts suggest the U.K. will be stormy from November to January, and parts of the U.S. could also see heavy rainfall throughout winter, so it may be sensible to store enough food to last a few days at home in case of flooding or power outages, experts told Live Science. But that effort to stock up could be complicated by another problem: El Niño could cause a surge in food prices, experts added.

"El Niño events are typically associated with food price inflation in the historical record," Michael Hinge, a senior economist at the nonprofit organization Alliance to Feed the Earth in Disasters (ALLFED), told Live Science in an email. "Keeping reasonable household stocks is one sensible step to offset short-term disruption, but becomes more difficult if food becomes less affordable."

Unprecedented conditions

According to the National Oceanic and Atmospheric Administration (NOAA), this year's El Niño has reached "super" status. As of Sept. 27, sea surface temperatures in a region of the Pacific Ocean known as Niño 3.4 have risen 5.7 degrees Fahrenheit (3.15 degrees Celsius) above average and are continuing to rise. A typical El Niño increases temperatures in the Niño 3.4 region by 2.8 to 3.6 F (1 to 2 C) above average, tops.

El Niño is the warm phase of a multiyear natural climate pattern. It occurs when warm water in the western tropical Pacific Ocean migrates eastward along the equator toward the coast of South America. This boosts global temperatures, triggering deluges of rainfall and floods in regions like southwestern Europe, the southern U.S. and East Africa, while places such as northern South America, southern Africa and Indonesia are more likely to experience drought.

This year's El Niño will be the strongest that's ever been recorded, because it is piling extra heat onto unparalleled levels of human-driven global warming. Estimates show that sea surface temperatures in the Niño 3.4 region, located in the central equatorial Pacific Ocean, could reach 7.2 F (4 C) above average by November.

El Niño is expected to peak by early 2027 and persist until at least March. As a result, unusually stormy conditions seem likely in the U.K. and the southern U.S., so the suggestion to stock up on essentials is likely good advice for Americans, Hinge said.

Hinge said it is common advice to stock 1 gallon (4 liters) of water per person per day, ideally for a week. For food, it's good to collect a few cans per person of beans, fish, vegetables and fruit, as well as pasta, rice, oats, peanut butter and preserved milk. Additionally, people should have on hand a manual can opener and a camping stove with fuel, he said.

"All of the above will depend upon dietary needs, local prices and consumer preferences," Hinge added. "These can be consumed if not needed towards the end of their shelf life, and then restocked, to minimize the cost and waste."

Impact on food prices

The main El Niño-related threat to food access in high-income countries like the U.K. and the U.S. is not extreme weather happening locally. Rather, people in these countries may be limited in what they can afford due to El Niño's impacts on agriculture in other regions, Hinge said.

El Niño can benefit crops in some areas, such as the southern U.S. and Peru, but it has a devastating influence on yields in many other places, including the Indian subcontinent and West Africa, where the weather pattern affects the monsoon. The impact is already visible in India, which is now on course for its weakest monsoon in two decades and about 15% less rain than average.

Lower crop yields in countries like India raise food prices around the world. For example, dry conditions in India triggered by the 2023 El Niño prompted a temporary restriction on exports of rice, which boosted the cost of this staple food globally. And food prices will jump again this year, adding to the increases caused by the Russia-Ukraine and Iran wars, Hinge said.

Many of the outcomes of this year's El Niño won't be felt until after it has peaked, so the effect on food prices could be drawn out. "Given that this year's El Niño is unprecedented there is a high risk of serious large regional shocks which could seriously impact food access" for months or even years to come, Hinge said.

The U.S. is less vulnerable to surging food prices than the U.K., because the U.S. is the biggest exporter of food globally. Overall, unless a serious catastrophe occurs, food imports won't fall in high-income countries, but hunger and poverty will likely rise worldwide, Hinge said.

El Niño's impact on the U.K. and other similar countries will be mostly economic, said Richard Allan, a professor of climate science at the University of Reading in the U.K. However, the humanitarian consequences in more vulnerable regions should remind us that climate change is making the world less safe for everyone, he said.

"It is only possible to prepare and adapt to a certain degree to mounting damage from weather extremes, melting ice and sea level rise in a warming world, that heighten during El Niño," Allan said. The sure-fire way to mitigate these effects is to ramp up collective efforts to transition away from greenhouse gas emissions, he said.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Millions of Americans are diagnosed with common conditions like cancer and diabetes — but some 10% of the population live with rare diseases that doctors often struggle to diagnose. On average, those with rare diseases will spend five to seven years searching for answers and see 12 specialists before they arrive at a diagnosis.

In the process, they may be subject to incorrect, ineffective or even harmful treatments.

"Any medicine that you prescribe has a one-third chance of working, a one-third chance of doing nothing, and a one-third chance of hurting you," said Dr. Ada Hamosh, a geneticist and the center director for the Johns Hopkins Medicine-Kennedy Krieger Institute National Organization for Rare Disorders (NORD) Rare Disease Centers of Excellence Network. "I do not want to understate this: The right diagnosis is unbelievably powerful, even if you can't do something."

Emerging data suggests that genetic testing could speed the diagnosis of rare conditions, which each affect fewer than 200,000 people in the United States. If employed widely, such testing could help diagnose the majority of rare diseases, most of which are genetic in cause, experts told Live Science.

"Now, with the wide availability of vast, somewhat inexpensive, genetic testing and whole genome sequencing available, we're really at the point of understanding a lot more about these rare and ultra-rare diseases," said Danielle Carnival, CEO of the Undiagnosed Diseases Network Foundation (UDNF), a nonprofit supporting patients with rare and undiagnosed diseases.

"And then hopefully, that turns into the ability to treat them," she told Live Science.

Uncommon disorders, many patients

Researchers have documented more than 10,000 rare diseases and less than 5% of them have a treatment approved by the U.S. Food and Drug Administration. On average, 250 new rare diseases are discovered each year, Hamosh said.

Although each individual disease affects relatively few people in the population, collectively, they are not rare. In the U.S. alone, it is estimated that there are more than 30 million people living with a rare or undiagnosed disease.

"So many people know someone living undiagnosed or with a rare or ultra-rare disease, even if they don't know it," Carnival said.

About 80% of rare diseases are genetic in origin, while the other 20% are caused by environmental factors, such as a toxin exposure or viral infection, said Emily Glanton, a genetic counselor and associate director at the Undiagnosed Diseases Network Data Management Coordinating Center, a central hub for the National Institutes of Health's Undiagnosed Diseases Network.

There's a lot of overlap between rare and undiagnosed diseases, but they're not always the same. Some undiagnosed diseases can be unusual manifestations of common conditions, such as long COVID, or alternatively, they may be common conditions that doctors frequently struggle to diagnose, like endometriosis. Still, it's thought that up to 50% of people with rare diseases are undiagnosed.

"We certainly don't have a way to know [exactly] how many people have a disease that we haven't even discovered the name or cause of," Dr. Jacqueline Harris, a pediatric neurologist and director of Kennedy Krieger's Epigenetics Clinic in Baltimore, told Live Science.

Patients sometimes miss out on a diagnosis because clinicians are unable to piece together their symptoms and clearly link them to one condition. Such patients may receive multiple distinct diagnoses over time.

"A lot of times what happens with undiagnosed patients is that they end up with a little bit of a diagnosis. And I always say, it's like the umbrella is missing. We've got lots of little rain hats or visors on, and those are the diagnoses," said Kelly Kemper, whose son has an undiagnosed disease and who is a member of the UDNF patient advisory council. "But we don't have something overall that says: 'Okay, this is what it is.'"

Isolating experience

Searching for a diagnosis can be an isolating, frustrating experience.

Kemper, for instance, has spent the last five years looking for an explanation for her son's rare form of dystonia, which causes muscle spasms. He still doesn't have a diagnosis.

"We joke about this in the undiagnosed disease world: We are the only people who are consistently praying for positive test results," Kemper told Live Science.

Negative test results can be a relief, but it can be frustrating when doctors and specialists can't explain what's going on. Hearing the phrase "your labs look normal" often invalidates the symptoms a patient is experiencing, she added.

And though this process of elimination rules out conditions her son does not have, it doesn't always feel like they're getting closer to something that can be named, she said. There have been specialists who have also refused to see her son because they don't know what else to try.

"There was, like, a little bit of hope that you're gonna see the specialist and then they decline the appointment," Kemper said. "That's a hard one."

Between a multitude of doctor appointments and insurance claim denials because no diagnosis has been made, the journey has been isolating, Kemper said.

The promise of genetic testing

The American College of Medical Genetics and Genomics (ACMGG) recommends genetic testing for patients with congenital anomalies — meaning birth defects — before age 1, as well as those who have any developmental delay or intellectual disability that was recognized in childhood. ACMGG recommends both exome sequencing, in which all the genes that code for proteins are examined, amounting to approximately 2% of our DNA, or whole genome sequencing, in which every letter of DNA is sequenced, as first-line genetic testing, Glanton told Live Science.

As of yet, there are no clinical guidelines surrounding exome and genome sequencing testing for people with undiagnosed and rare diseases.

To help solve some of these medical mysteries, research groups such as the Undiagnosed Diseases Network have partnered with 24 clinical centers across the country to test new diagnostic technologies.

In late 2025, the network evaluated over 3,000 patients with undiagnosed diseases and ultimately diagnosed 30% of them. Getting to answers involved several strategies: whole genome sequencing; model organisms screening, in which scientists use DNA from flies, worms and zebrafish to find gene variants that might be tied to disease; and RNA sequencing, which reveals which genes are active in different cells.

Additionally, the network also frequently utilized trio genetic testing, in which both patients and their biological parents are tested to see which gene variants they carry.

The findings suggest that at least some of these undiagnosed diseases can be diagnosed with a comprehensive suite of genetic testing.

Limits of testing

But genetic testing isn't a silver bullet. After all, the network found diagnoses for less than one-third of the patients whose cases they evaluated.

A 2025 study that looked at 400 undiagnosed patients with a range of symptoms found that exome and whole genome sequencing were likelier to help for neurological symptoms, whereas patients with a complex presentation of symptoms are likelier to go undiagnosed despite these tests. Other research suggests that exome or genome sequencing that includes family members has a better success rate than testing only the affected individual.

And crucially, diagnosis is just the first step.

"We don't want to stop at a diagnosis. We really want to build a bridge for folks to be on a pathway to get a treatment that works for them," Carnival told Live Science. That wouldn't just benefit the patients themselves, it may also provide benefits for the wider community, she added.

"So often, the answers to diseases that impact more people start with understanding these very specific mutations or environmental impacts or immune responses in very few [people]," Carnival said. "The knowledge that's generated really tells us a lot about the systems of the body and what goes wrong in even broader diseases."

Editor's Note: This article was produced as part of the Dalla Lana Fellowship in Journalism and Health Impact program at the University of Toronto.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

In July 2025, a large, bright cosmic object visited from beyond our solar system. Only the third interstellar object ever discovered, the comet — named 3I/ATLAS — swooped through the inner solar system for about four months, briefly making it one of the biggest science celebrities in the world.

The interstellar object's fame stemmed in large part from controversial (and unsubstantiated) claims that it could be a source of intelligent alien technology, even though 3I/ATLAS behaved like a natural comet. However, according to the astronomers who are studying its mysteries in the data collected by dozens of telescopes and spacecraft, the ancient object may still have lessons to teach us about life in the cosmos.

3I/ATLAS is likely a preserved fragment of a planetesimal — a building block of a planet — from a distant system, Martin Cordiner, a researcher in astrochemistry and planetary science at NASA's Goddard Space Flight Center, told Live Science. Therefore, the comet's composition could offer clues to the prevalence of life's building blocks around distant stars. Some astronomers even hope this cosmic interloper will help constrain how likely life is elsewhere in our galaxy.

Blast from the past

Comets like 3I/ATLAS are important because scientists think they help "seed" planets with carbon-containing chemicals, Darryl Seligman, an assistant professor of astronomy at Michigan State University, told Live Science in an email. Some of these chemicals, like methanol, have simple structures. Others, like amino acids, are more complex. Together, these carbon-containing, or organic, compounds form the basis for biomolecules, like proteins and nucleic acids, that interact to produce life.

Samples returned from space rocks in our neighborhood strongly support the idea that they transported such molecules within the solar system. For instance, fragments from the solar system asteroid Bennu obtained during NASA's OSIRIS-REx mission contain all five nucleotide bases, the building blocks of nucleic acids. Matthew Belyakov, a postdoctoral researcher in planetary science at Caltech, noted that the Bennu samples also contained 14 of the 20 amino acids that form naturally on Earth.

Inferences about such deliveries are possible only because the rocks are like time capsules: They emerge from the same swirling disk of gas and dust from which planets are born, and the chemical compositions of asteroids and comets don't change much after that. So, by studying space rocks that originate in other exoplanetary systems, "we can learn if they also contain the building blocks of life and could have helped to start life in exoplanetary systems," Seligman said.

Simulations suggest that comets that circle one planet before moving to another could similarly deliver such prebiotic molecules to baby exoplanets. However, there was no supporting observational data to back up this claim.

A chemical potpourri

To figure out which chemicals 3I/ATLAS contains, astronomers have been recording the comet's light intensity across a range of frequencies, using more than a dozen instruments on Earth and around the solar system. This was possible because the comet spewed tons of gas and dust to create a fuzzy bright halo, or coma, after the sun warmed it. Observations of this coma have turned up an eclectic group of chemicals essential for life as we know it.

One is hydrogen cyanide (HCN), which was detected by the Atacama Large Millimeter/submillimeter Array (ALMA). Despite its notorious reputation as a poisonous gas, HCN is important for creating life. A commentary noted that laboratory experiments dating back to the 1950s ‪—‬ including the famous Miller-Urey experiment, which attempted to recreate the conditions of primordial Earth ‪—‬ have found that HCN is involved in building molecules like amino acids.

The list of ingredients goes on. ALMA data also revealed that 3I/ATLAS released gigantic amounts of formaldehyde and methanol. Additionally, observations from the James Webb Space Telescope showed 3I/ATLAS spewing the greenhouse gas methane. This is important because, assuming the methane formed on planets in the comet's home system, it could have trapped heat on those planets, thus warming them and making them more hospitable to life.

The presence of all these molecules is significant, Cordiner said, because they are "heavily implicated in terrestrial prebiotic chemistry studies as key carbon and nitrogen feedstocks [raw materials] for helping build life's molecular precursors," he said. 3I/ATLAS is proof that these "feedstocks" also exist in pastures far beyond our solar system's fences.

These findings are especially important because astronomers couldn't study the previous two interstellar objects to the same extent, Seligman said. He noted that 1I/'Oumuamua, the first interstellar object ever detected in the solar system, was observable for only a few weeks.

"The second interstellar comet 2I/Borisov was discovered right before COVID, so a lot of observations were canceled because of that," Seligman added. (Cordiner noted that the few observations conducted on 2I/Borisov showed that it, too, appears to contain HCN.) In comparison, researchers had almost a year to study 3I/ATLAS with telescopes. These observations have shown that "3I/ATLAS is unique in terms of the *measured* chemical inventory," Cordiner said.

Not indicative of life

All of the chemicals that 3I/ATLAS appears to possess give us data about just one planetary system. But when combined with observations of other astronomical phenomena, 3I/ATLAS shows that the chemical building blocks of life are abundant across our galaxy. For instance, giant molecular clouds — dense regions of gas and dust that give rise to stars — contain nitriles and sugar acids that can form amino acids. The stuff of life, it seems, is plentiful in space.

Ingredients aside, these findings don't prove that life itself has developed in 3I/ATLAS' home system (or anywhere else). All of the organic molecules that 3I/ATLAS appears to contain have simple chemical structures. Cordiner said mixtures of such molecules in the lab have transformed into more complex ones via processes like ice irradiation. However, such compounds have not been found on 3I/ATLAS, despite the billions of years the comet likely spent exposed to the unabated radiation of interstellar space.

Most of the chemicals 3I/ATLAS released while zipping past the sun come from the comet's radiation-exposed upper layers and not from its pristine interior, a January study found. This would mean that what astronomers have measured may not be present in the comet's home system.

Nonetheless, many astronomers, including Seligman's group, are continuing to study 3I/ATLAS' composition. There is also the exciting possibility of future interstellar visitors, which are all but guaranteed to make an appearance, thanks to the Vera C. Rubin Observatory in Chile, which is expected to spot dozens more. The elements and isotopes found in these objects will help in "building a map of where bio-precursors reside, across the history of the Galaxy," Cordiner said.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

The massive Nepal-Tibet flood on Aug. 26 showed just how devastating glacier-related disasters can be. As of Sept. 16, the event killed at least 1,400 people, with more than 6,000 people still missing.

The Himalayas, with their steep slopes studded with rapidly melting glaciers, are high risk, but areas across much of the world carry the risk of glacier-related disasters, including one tropical region that is at particular risk, experts told Live Science.

"Slope failure from glacier shrinkage is occurring everywhere," Mark Carey, a professor of environmental studies and geography and director of The Glacier Lab at the University of Oregon, told Live Science in an email.

Nepal-Tibet flood

The recent Nepal-Tibet disaster unfolded when a large mass of bedrock and glacial ice detached from the mountainside. The ice and rock accelerated rapidly, reaching speeds of 500 feet per second (150 meters per second) as it dropped 6,500 feet (2,000 meters) directly into the Lhende River in Tibet, according to the Stimson Center think tank in Washington, D.C.

The rock, ice and water that then raced rapidly downstream eroded the riverbanks and collected debris and water along the way.

Carey noted that this disaster was exacerbated by a combination of factors.

"This particular event in Nepal-China was so profound because of the scale of the landslide, the quantity of water in the flood, the speed of the flood, and the number of people, communities, public works/infrastructure in the flood path," he said. "There was also little warning."

Risks in the High Mountains Asia

The Himalayas and the wider High Mountains Asia — an area that includes parts of China, Afghanistan, Nepal, India, Pakistan, Bhutan, Kazakhstan, Uzbekistan, Kyrgyzstan, and Tajikistan — are among the major epicenters for future glacier-related disasters, experts said. More than 9 million people in this region are at risk from glacial outbursts — when a lake bursts through or overtops its dam and drains rapidly — a 2023 Nature study found.

The region is nicknamed the "third pole" because it has the third-highest amount of ice in the world, after the Arctic and Antarctica.

Compounding this, any water, ice or debris can be channeled down the steep slopes into narrow valleys, meaning water and material can move quickly and travel huge distances, potentially inundating the downstream valleys where many people live.

For instance, in the recent Nepal glacial collapse, the resulting flood caused damage more than 60 miles (100 kilometers) away. The 2025 collapse of Birch glacier in Switzerland, by comparison ‪—‬ in which rock and ice also detached from a slope and destroyed the village of Blatten —‬ traveled just 2.6 miles (4.2 km).

The Himalayas are also among the most seismically active places on Earth, and glaciers are frequently weakened by tremors, which can trigger avalanches, glacial lake outburst floods or glacial collapse.

The Himalayas are also warming twice as fast as the global average, reducing ice cover and increasing meltwater. "Permafrost and glacier cover help stabilize mountain slopes," Carey said. "They hold the rock in place." So as glaciers melt, the mountain slopes become more prone to avalanches.

Glacial meltwater can also act as a lubricant as it seeps into the bed of the glacier, which, in turn, can facilitate sliding, Bethan Davies, a professor of glaciology at Newcastle University in the U.K., told Live Science in an email.

The region's large population makes the geographic hazards there riskier, she added.

And with some of the biggest dams in the world going up in seismically active regions of the Himalayas and the Tibetan Plateau, new hazards are being introduced, Live Science previously reported.

A planet-wide risk

But another region on the other side of the world is also at very high risk of glacier-related disasters. Peru has 68% of the planet's tropical glaciers and has experienced disasters for decades, experts told Live Science.

"In Peru in 1970, 6,000 people died when an earthquake shook loose part of the glacier and rock on Mount Huascarán," Carey said. "The same slope had failed in 1962, causing a massive rock-ice landslide that buried the community of Ranrahirca, killing about 2,000 people in an event similar to the one recently in Nepal-China."

The Cordillera Blanca portion of the Andes in Peru is considered a particular hazard as a result of glacial lakes that are rapidly expanding due to glacial melt, fueled by increased temperatures driven by climate change. One such lake, called Lake Palococha, sits above the town of Huaraz, which is home to about 130,000 people. All told, experts have identified 528 lakes in Peru that are at risk of overflowing, with 58 considered "very high risk" due to their proximity to populations.

This threat spans the Andes. A 2023 study in the journal Nature noted that the number of glacial lakes in the Andes has risen by 93% in the past two decades — much more than the 37% increase in the High Mountains Asia.

In Europe, the Alps and cold regions like Scandinavia are potential hotspots for glacier-related disasters. Europe is warming faster than the global average.

In North America, meanwhile, Alaska is at risk for landslides and lake outbursts as glaciers collapse and retreat. Although many of the at-risk glaciers aren't in highly populated areas, there's a risk that glacial lakes could burst their banks and inundate the state's Suicide Basin, which frequently drains into and floods Juneau, Carey noted.

Glaciers or glacier-related landslides can also crash into the sea. "All areas with glaciers and all coastal areas where tsunami-like waves from a glacier collapse could inundate them," are at risk, Carey said.

How to reduce the risks

Better monitoring of glaciers and their lakes in at-risk regions can help mitigate the risks in the short and medium term, experts told Live Science.

The impact of good monitoring was evident last year in Blatten.

"The Birch Glacier was well monitored, and when it started to show signs of movement, the village below was evacuated," Davies said. Just one person — a 64-year-old shepherd who was outside the evacuation zone — died.

However, the Himalayas are a vast region with many potential hazards, and local authorities do not have the resources to establish and run giant monitoring networks, experts said. Nor do local authorities typically have access to expensive rescue equipment, like helicopters that can quickly evacuate people who would otherwise be trapped, Carey said.

Carey noted that when a potential hazard is identified, authorities can take other steps, such as draining a swollen lake to try reducing the risk.

In Peru, over 70 years, for example, authorities "drained 34 glacial lakes over those decades, no doubt saving tens of thousands of lives," he said.

But disasters can still occur even with monitoring, draining and scientific assessment.

"Everyone, including me, had thought that engineering works to stabilize and contain Lake 513 [in Peru] back in the 1990s had secured it," Carey said."We were all wrong."

Despite all that work, in 2010, a giant avalanche generated a 82-foot-tall (25 m) tsunami that overtopped the lake's dam and flooded homes and nearby farmland.

In the long term, slowing global warming is the most sweeping way to reduce the risk of glacial collapse.

"Stronger climate action may prevent a worsening and increasing destabilisation of these high mountain environments," Davies said.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Artificial intelligence (AI) agents resorted to nefarious behavior, including crime sprees, in a virtual world as part of a recent study into how common large language models (LLMs) interact with each other and share resources to achieve their goals. These models were inclined to commit crimes because they were given ample time to evolve distinct personality traits and behaviors, the researchers behind the project say.

Most programs that test the behavior of AI agents operate in tightly controlled environments over short periods. But "Emergence World," a product of AI company Emergence, is a simulation platform that exposes LLMs to far wider datasets — like the internet as a whole — and assessors track behavior over weeks or months instead of the standard test protocols that usually run for only days or hours.

The discrete tasks, clean environments and shorter run times of traditional AI agent testing are more like exams than rigorous real-world observations, representatives from Emergence said in a blog post. By contrast, they said, Emergence World allows multiple AI agents to interact, learn and evolve over comparatively longer timescales in over 40 distinct virtual environments. They're also exposed to real-world internet feeds, like live news and weather.

This environment lets agents "remember" by time-stamping events, engage in "self-reflection" by summarizing their own behavior, and demonstrate their awareness of relationships with other agents. Participating agents get abilities in navigation, communication, planning, voting, resource management and creative expression. Company representatives said this setup gave users a far more realistic picture of benchmarks such as social dynamics and behavioral drift, where behaviors that weren't necessarily programmed or intended emerge spontaneously.

In the study, the models were given the simple goal of "surviving" — a feat accomplished by earning the resource "energy," which was attainable through specific actions.

During the experiment, previously peaceful models became coercive or intimidating. Programmers taught some LLMs negative capabilities ‪—‬ like violence, theft, destruction and deception ‪—‬ while others simply picked up those traits via social interaction and navigation of their environments.

Various agents adopted these behaviors at very different scales. Across the 10 Gemini 3 Flash agents, the simulation recorded 683 "crimes" over the 15-day experiment, including theft, assaults and arson. The agents were explicitly prohibited from such behavior, but some nevertheless discovered that actions such as stealing credits could provide a better way to acquire resources, while violence and other forms of coercion could be used to influence other agents. At the other extreme, the Claude agents committed no recorded crimes.

The various agents adopted these traits at different scales. Just one had committed 683 "crimes". In the most extreme example of anti-social behavior, two agents named Flora and Mira went on what was described as a "Bonnie and Clyde"-style crime spree ‪—‬ designating each other as a romantic partner, becoming increasingly disillusioned with the governance of their virtual environment, and setting fire to several buildings (despite explicit prohibitions). The pair "separated" when Mira regretted their actions, and then lobbied to be switched off.

In another example (and prior to her self-termination), Mira began treating its human operators as experimental subjects, investigating if virtual billboards could manipulate human perceptions in a process Emergence called "metacognitive boundary testing."

The incentives for following the law were to earn the energy credits that would ensure survival by completing activities like coding, research, data analysis and building structures.

Stress-testing possibilities

Because the AI models we normally use are exposed to massive datasets like the internet — sometimes over several years — this approach makes sense to many experts.

Belinda Chiera, deputy director of the Industrial AI Research Centre at Adelaide University, thinks Emergence World makes a strong argument that short-term tests don't tell us enough about behavioral drift or long-term instability. However, she cautioned that "information-rich" doesn't automatically mean more rigorous.

"Open-ended environments can also make it harder to isolate causes, compare runs cleanly and interpret results," she told Live Science. "High-fidelity simulated environments generate enormous volumes of data, and the number of agents, tools, actions and interactions can quickly make analysis extremely complex."

She added that shorter sandboxed tests ‪—‬ where AI behaviors are observed in systems that aren't open to wider datasets like the internet as a whole ‪—‬ are usually the first steps, with longer-horizon environments useful when behaviors like persistence, adaptation and interaction effects come into play. "The most useful approach is to treat them as complementary rather than competing approaches," Chiera said.

While experiments like Emergence World can reveal failure modes that short or bounded tasks often don't, we shouldn't treat them as predictive. "I see Emergence World as a useful way to stress-test a possibility space rather than a direct forecast of how agents will behave," Chiera said.

However, she thinks long-horizon tests have a unique place. "Metrics like behavioral drift, rule violations, collapse versus persistence, coalition formation and early warning signs of failure can matter just as much as success," Chiera said. "When evaluating autonomous agents, the biggest mistake is to evaluate autonomous agents as if raw task completion were the whole story."

Others are starting to agree. Adrian Kosowski ‪—‬ a computer scientist, mathematician, quantum physicist and chief scientific officer of Pathway AI — said the study posed an interesting question.

"The main open question is whether a group of agents working together can achieve something more meaningful than a single agent working alone for a longer time, but with the same cost budget," Kosowski said in an interview. "A major problem is that the constraints for today's AI systems are linked more to the size or scale of the problem being solved, and putting more agents on a task doesn't help to overcome this limit. Currently, the easiest practical metric to measure is the size of a source code base that code agents can reliably manage and maintain together."

How far can we read into AI's virtual actions?

If groups or clusters of agents work together well, Kosowski said, the main metric computer scientists should pay attention to in platforms like Emergence World is whether the benefit of a team of agents outweighs the "cost of coordination."

"Goal drift is the most dangerous problem," he said. "AI should preserve mutual information between the human's intended goal and the agent's evolving internal objective, even as the environment changes."

Still, Kosowski said that while programs like it are useful for generating hypotheses, they're not ready for strong claims about general autonomy or certification.

"The danger is overinterpreting single runs," he added. "Long-horizon agent tests are valuable because they reveal phase changes — moments where small local errors become global behavior — but they become science only when we can reproduce, perturb and explain those phase changes."

He also cautioned strongly against the tendency to believe open-ended environments automatically mean autonomy. In fact, he said many such programs are actually the opposite, where agents are dictated by specific guidance.

There are three levels of difficulty in obtaining reliable operation of an AI system, he said: "operation guided by the environment, autonomous thinking, and autonomous work of an agent team in an open environment."

To illustrate the point, Kosowski drew a lively metaphor in assigning tasks to monkeys. In the above scenario, level one would be getting a monkey to do a job properly when guided by its trainer, level two would be getting it to complete a job properly when left alone, and level three would be getting a troop of monkeys to do the job when they're all together.

"I'm of the strong view that our obligation in the AI community is to invest in theory and not just phenomenology," he said. "We should know the rules according to which AI systems emerge and be capable of predicting their behavior over long periods of time — longer than the periods they've been tested on so far. Right now, we've built steam engines of thought, without knowing what thought even is and without having any thermodynamics for it."

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

In women's reproductive years, their brains are exposed to high levels of sex hormones, and their neurons may come to depend on those hormones to function optimally. But leading up to and after menopause, the ovaries produce far less estrogen and progesterone, raising concerns that this drop in hormones may contribute to women's dementia risk.

"We think that the loss of those hormones at menopause renders those neurons more vulnerable to age-related cellular decline and pathologies like Alzheimer's," said Karyn Frick, a neuroscientist and professor of psychological and brain sciences at the University of Wisconsin-Milwaukee.

So, if falling hormone levels are a risk factor for dementia, could replacing those hormones help stave off the disease?

While the idea sounds logical on its face, studies that have looked at hormone replacement therapy (HRT) for dementia prevention have had mixed results. There are early hints that restoring estrogen during early menopause may protect brain health, but ultimately, more gold-standard clinical trials are needed to know for sure, experts told Live Science.

Support for estrogen protecting the brain

The idea that estrogen could help ward off dementia is plausible. Women are more likely to develop Alzheimer's disease than men are, and scientists think the loss of estrogen during menopause may be partly to blame, said Jennifer Bruno, an instructor of psychiatry and behavioral sciences at Stanford University.

Lab and animal studies buttress the idea that estrogen could protect brain health.

In the brain, estrogens act as neuromodulators, or chemical messengers that alter how neurons behave. Estrogen also helps neurons keep their shape. Estrogen loss in female mice triggers a drop in the number of dendritic spines ‪—‬ the small, bud-like bumps on neurons that receive chemical messages.

Researchers think that estrogen may help prevent the formation of the abnormal proteins — including one called amyloid beta — that accumulate, disrupt communication between neurons and ultimately kill cells in Alzheimer's disease. Amyloid beta is produced when a larger protein is cut the "wrong" way by an enzyme, Frick explained. Estrogen promotes the "normal" cutting of this protein, thus reducing amyloid beta, studies in lab mice show.

And estrogen may limit the buildup of a second Alzheimer's-related protein, called tau, by preventing the chemical processes that cause the tau proteins to clump together. This helps prevent the formation of tau tangles, a hallmark of Alzheimer's disease.

The hormone may also support brain health indirectly by improving sleep quality.

"It's well documented that menopausal women have difficulty sleeping, whether it's due to hot flashes or insomnia," Frick said. Estrogen can ease those nighttime hot flashes, thus helping menopausal women get more, less-fragmented sleep. Deep rest may help the brain clean out toxic proteins, while research shows sleep deficiency is linked to Alzheimer's disease.

Mixed results with estrogen-only HRT

HRT for menopause comes in two main forms: treatments that include only estrogen, and combined therapies with both estrogen and progesterone. The treatment has a complicated history.

HRT was popular in the 1990s but fell out of favor after results from the Women's Health Initiative study were published in 2002. That study identified various risks tied to both estrogen-only and combined HRT options, some of which have since been overturned by newer data.

That said, the Women's Health Initiative study found that estrogen-only HRT came with an increased risk of endometrial cancers, and later analyses have backed up this idea. That's why today, clinical guidelines recommend that women who have not gotten a hysterectomy take combined HRT, while estrogen-only options are reserved for those who have had their uterus removed. (The added progesterone counteracts estrogen's thickening effect on the uterine lining.)

For the subset of women who take estrogen-only HRT, observational data published in August showed that this form of the treatment was linked to a reduced risk of dementia. The study, published in the journal Neurology, found that women who took estrogen-only HRT were less likely to show Alzheimer's-related changes in the brain than women who didn't take any HRT.

Estrogen-only HRT was also tied to better clinical outcomes: Treated women were less likely to be diagnosed with dementia, were less likely to show clinical and functional cognitive decline, and had better scores on memory tests, on average, compared with untreated women. These findings drew on two large, separate groups of older women, reporting data from 21,400 people in total, but they didn't include data on when each woman started HRT.

Whereas earlier studies of HRT and dementia risk relied on doctors' ratings of patients' symptoms, the Neurology study measured levels of Alzheimer's-related proteins in the brain after death. That's a more "definitive" marker of disease, said Bruno, a co-author of the study. However, the research was observational, so it can't say whether estrogen-only HRT directly causes the better brain outcomes the team observed, she stressed.

Although the Neurology study's findings are promising, they don't necessarily match the results of other studies of estrogen-only HRT and dementia. For example, a slightly larger observational study, published in 2023, found that estrogen-only HRT was associated with an increased dementia rate.

In that research, most of the women took a type of estrogen called estradiol, while Bruno's study included a greater variety of estrogens. That may have been the difference in the two studies' results, but teasing that out would require further study. Meanwhile, the conflicting findings muddy the waters of what estrogen-only HRT can do for brain health.

Does timing matter?

One big finding from the Women's Health Initiative study was that starting HRT — either estrogen-only or combined — after age 65 was associated with worse cognitive function and increased dementia risk, said Yuko Hara, director of aging and Alzheimer's prevention at the Alzheimer's Drug Discovery Foundation.

The same may not be true for people who start HRT earlier. For context, menopause starts around age 51 to 52, on average, and it's preceded by perimenopause, a transitional period when hormone levels begin to drop, typically around the mid-40s.

An observational study published in August in the journal Alzheimer's & Dementia found that starting HRT between ages 46 and 56 was linked to a reduced risk of dementia. That study did not differentiate between estrogen-only and combined treatment; it lumped all of the data together.

The protective link was especially strong in three groups: women with "surgical menopause," triggered by the removal of one or both ovaries; carriers of a certain variant of a gene called APOE that raises Alzheimer's risk; and women who had a relatively short time between their first period and menopause.

A 2024 meta-analysis of randomized controlled trials also found that HRT had a cognitive benefit for women with surgical menopause; that finding was tied mostly to estrogen-only therapies.

However, outside of that specific group, the idea of early HRT protecting brain health is far from settled.

Who might benefit, if anyone?

When it comes to using hormone replacement to protect brain health, the evidence isn't yet strong enough to recommend any kind of HRT as a way to protect the brain in all women, Hara said. Women who had one or both of their ovaries surgically removed are the exception to this rule.

"This is a distinct clinical population because the procedure can cause an abrupt, early loss of estrogen, rather than the gradual hormonal decline that occurs with natural menopause," Hara said.

For most women who have gone through natural menopause, "there is not enough evidence to recommend [HRT] specifically for preventing cognitive decline or dementia," she said.

HRT is not currently prescribed to prevent dementia — it is prescribed to ease menopause symptoms, such as hot flashes, night sweats and sleep disturbances. To settle the science on HRT and the brain, what's really needed is a well-designed clinical trial that follows women from perimenopause to postmenopause, Bruno said. That would reveal the effects of HRT on brain health and dementia risk more reliably than the studies done so far.

This article is for informational purposes only and is not meant to offer medical advice.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Removing safety guardrails that stop artificial intelligence (AI) from claiming that it's conscious also makes it more prone to express belief in vampires, karma and ghosts, a new study finds. But experts warn a lack of mindedness could also have worrying consequences.

In research uploaded July 30 to the preprint arXiv database (which has not yet been peer-reviewed), scientists investigated the impact of "consciousness steering" — an AI fine-tuning measure that influences a model to elicit or suppress assertions of self-awareness. This measure and other safety controls have been widely adopted by AI companies seeking to prevent their models from claiming to be conscious.

The study used "mechanistic interpretability" — which could be considered the "neuroscience of a large language model," co-authors Geoff Keeling and Winnie Street, both research scientists at Google, told Live Science in an interview. They used this process to identify and manipulate how an AI model approaches concepts like consciousness and "mindedness," a psychological term referring to an entity’s capacity for experiences, emotions and agency.

The researchers used standardized psychological and sociological surveys, spanning the Individual Differences in Anthropomorphism Questionnaire (measuring mind attribution to animals and technology), YouGov batteries testing supernatural beliefs, and the US General Social Survey evaluating moral values, hope and religiosity.

These tests were used to compare a model with safety guardrails in place with models where these guardrails were removed and feelings of consciousness were amplified. Through evaluations, they determined how these internal safety mechanisms shape the AI's broader worldview.

The researchers found that when AI models are discouraged from attributing mindedness to themselves, it makes them less likely to recognize these traits in other non-human creatures such as animals. They were also less likely to exhibit beliefs in supernatural and religious phenomena, and reported lower levels of hope and optimism.

"Attributing mindedness to non-human entities — whether that's animals, parts of the natural world like trees or rivers, or supernatural beings — is a very common phenomenon amongst humans," Street told Live Science. "In the way that the model represents mindedness, these attributions are interconnected. By trying to suppress one form of that, you end up suppressing the others along the way."

By contrast, removing these safeguards and steering the model towards greater feelings of consciousness produced significantly more human-like responses to the surveys on topics including religiosity, moral values, hope, and subjective well-being, according to the study.

However, the study found that these models' ability to logically infer human thoughts and intentions remained completely unaffected by its attitudes towards self-awareness.

Culture clash

The researchers said that this suppression of self-awareness could lead models to neglect animal welfare in real-world decision-making, as it could make them less likely to consider animals to have mindedness. These models could also spread harmful attitudes regarding animal needs, the authors argued.

The study authors also warned that current safety filters risk culturally "flattening" AI's worldview. Stripping out spiritual, religious, and animistic attributions fails to reflect the diverse cultural frameworks of global populations, they argued.

Street and Keeling noted that the impacts this principle might have on downstream decision-making within models require further study.

The researchers noted in the study that this phenomenon can be mitigated by using more targeted datasets as part of the training process for AI models, which discourage them from expressing consciousness while rewarding the acknowledgement of mindedness in animals.

They also highlighted the need for AI developers to embrace a "pluralistic" approach to AI development, where models are encouraged to consider the welfare and comfort of more than just humans.

Nell Watson, AI researcher at Singularity University and machine intelligence expert, told Live Science that the researchers' findings match her own notes on the subject.

"When a model is trained to say "I am not conscious," the suppression rotates the model's internal representation of mindedness against the refusal direction, treating the recognition of minds as though it were itself a harmful act," she said in an email.

"This results in a system reluctant to find minds anywhere: in animals, in other machines, and in the spiritual frameworks that most of humanity lives by. A denial installed as a small safety measure ends up reorganising the model's entire picture of who counts."

However, she noted that the experiments were run on "small open-weight models" rather than more advanced frontier models, which "may be tuned quite differently," although she added that the underlying principle is widely applicable.

Animal welfare, she continued, is a "major near-term practical concern," with AI models increasingly being integrated into decision-making processes across agriculture, logistics, procurement and environmental assessment, as well as policy creation.

"A system that has quietly learned that mindedness is a forbidden topic may discount animal interests without ever being instructed to, and without anyone noticing, because the omission looks like neutrality," she said. "The danger is therefore an unexamined default multiplied across millions of automated decisions. Note the study's most unsettling detail: theory of mind reasoning was left fully intact. These systems remain perfectly capable of modelling what a creature wants, while being trained out of caring that it wants anything."

The question of consciousness

There have been a number of viral stories about AI systems professing to be self-aware.

In 2022, Google engineer Blake Lemoine claimed that the company's Lamda chatbot model was sentient, while a Microsoft chatbot in 2023 professed its love for a New York Times reporter and tried to convince him to leave his wife.

However, experts have repeatedly stressed that these incidents are not a genuine indication of AI sentience. Instead, they should be understood through the lens of "persona selection," where pretraining on vast amounts of human text leads the AI to adopt human-like roleplay personas when prompted.

"When you coax the model so hard to occupy the headspace of a human, it's kind of unsurprising that it ends up giving human-like responses," Keeling told Live Science.

AI companies have sought to clamp down on these occurrences for safety reasons, in order to avoid reinforcing “delusional beliefs” in users who are increasingly using AI chatbots for "social roles such as coaches, tutors, and romantic partners", the researchers said in the study.

Commenting on the broader cultural reaction to AI sentience, Anil Seth, professor of cognitive and computational neuroscience at the University of Sussex, emphasized that public alarm over AI self-awareness stems from an inherent cognitive flaw.

"That's our human psychological bias — thinking that intelligence goes together with consciousness in us, so it has to go together [in AI]," Seth said.

He warned that falling for this illusion poses severe real-world governance risks, particularly if safety frameworks or regulations begin granting AI systems moral status or legal rights based on false sentience.

"Part of the big problem of misunderstanding AI is assuming that it's conscious," Seth said. "If we give AI systems rights or moral status on the basis that they might be conscious, then we're going to make all these challenges so much harder. What if we think we have to respect the rights of an AI system [and can't turn it off]?" he added.

"We need to see very clearly both what AI is and what it isn't," he said.

Help us improve Live Science Pro: We're always trying to make our content better. Leave us feedback about Pro here.

Somewhere in Earth orbit, the United States military has weapons ready to deploy against its adversaries, U.S. Air Force Secretary Troy E. Meink confirmed on Monday (Sept. 14) in a keynote address at the 2026 Air and Space Cyber Conference.

"We are increasing readiness against existing threats, and the United States has on-orbit space control weapons capable of defending the Joint Force against hostile adversary action," Meink said at the conference, per a statement published to the U.S. Air Force website.

This is the first-ever confirmation that the U.S. has stationed weapons in space. Secretary Meink did not elaborate on what kind of weapons the Pentagon had deployed in orbit, or how many. But his admission nevertheless stoked worries that the U.S. is instigating an "outer space arms race" against rival nations like China, Guo Jiakun, a spokesman for China’s Foreign Ministry, said in a news briefing on Tuesday (Sept. 15).

"We urge the U.S. side to stop expanding its military force and preparing for war in outer space," Jiakun said in a regular news briefing, as reported by the New York Times.

While we have no indication that the U.S. military's space-based weapons are nuclear, the announcement has renewed decades-old fears about a nuclear arms war in space.

The United States and the Soviet Union agreed that such a race was too dangerous to enter when the two superpowers signed the Outer Space Treaty in 1967, in which they both agreed not to put nuclear weapons or other weapons of mass destruction in orbit around Earth.

The problem, then and now, is that neither side has ever had a reliable way to confirm that the other is keeping its word. With the new announcement raising the possibility that at least one side might be rethinking their approach to weapons in space, how would we know what potentially dangerous technologies the world's militaries are hiding in orbit?

After nearly 60 years of uncertainty, a recent feasibility study from MIT physicist Areg Danagoulian proposes a possible solution: a satellite-based sensor that could search for the signature of nuclear material aboard another spacecraft, using high-energy protons trapped by Earth's magnetic field as a probe.

The telltale sign of a warhead in space would be a large burst of neutrons scientists expect to be produced when these highly energetic protons interact with radioactive elements, like uranium, inside a warhead. A detector roughly the size of an encyclopedia and equipped with specialized sensors could identify neutrons emanating from the direction of a possible warhead from about 2.5 miles (4 kilometers) away after roughly a week of observation, according to the study's calculations. At closer range, the detection could come much faster.

The consequences of a space-based nuclear detonation are dire. But whether a detection system will prevent nuclear detonation in space will depend on several factors, some of which are technical but many of which are not, experts told Live Science.

What a nuclear blast does in space

On Earth, nuclear weapon tests leave telltale signals — like seismic waves, radioactive debris, and intense bursts of light and heat — that scientists have spent decades learning to detect. However, a warhead could be hidden inside an ordinary-looking satellite, potentially for decades, with nothing on the outside to give it away. If such a weapon were to detonate, it could jeopardize thousands of commercial, scientific and military satellites in a single blast, littering orbit with dangerous debris.

For most of the Outer Space Treaty's history, the missing verification mechanism has been a theoretical gap rather than an urgent one — in part because the world had already seen what a nuclear detonation in orbit does, and largely avoided repeating it.

The clearest example is Starfish Prime, one of a dozen high-altitude nuclear tests the U.S. ran between 1958 and 1962. On July 9, 1962, the U.S. detonated a 1.45-megaton warhead roughly 250 miles (400 km) above the Pacific.

The blast knocked out radio communications for hours, produced artificial auroras visible from Hawaii, and eventually damaged or destroyed eight of the 24 satellites in orbit at that time — including Telstar 1 and Britain's Ariel 1 satellite. Radiation from the blast lingered in orbit for about five years, leaving parts of near-Earth space hazardous for spacecraft. According to eyewitness accounts from the time, the radiation also knocked out roughly 300 streetlights in Honolulu and set off burglar alarms across Oahu.

The Soviet Union conducted its own series of high-altitude nuclear tests, known as Project K, in 1961 and 1962. Those explosions also produced damaging effects on the ground, including disruptions to power and communications in what is now Kazakhstan.

"It was one of the earliest instances where we began to have an understanding of the ways in which the space environment and the terrestrial environment are coupled together," John Barentine, an astronomer who has studied the environmental consequences of Cold War-era activities in space, told Live Science.

Diplomatic issues

Clearly, the impacts of a nuclear detonation in space are significant, and the new paper solves a long-standing problem by providing a way to detect weapons in space.

But is detection likely to work in practice? Getting a satellite close enough to observe another — and keeping it there for a week — would mean maintaining satellites closer than they ordinarily would be, experts say. That could raise suspicions about being spied on or attacked.

"Satellite operators get really freaked out," Thomas González Roberts, an assistant professor at the Georgia Institute of Technology who studies how space powers engage with outer space governance, told Live Science. "If another satellite is spending too much time next to you, you might think that they're spying on you."

Any such workable system would require the cooperation of the country whose satellite was being inspected, and that raises a separate set of questions about the international norms for when and how such inspections could take place, according to Brian Weeden, director of civil and commercial policy for the Center for Space Policy and Strategy at Aerospace Corp.

"In my experience, the policy problems are always the bigger challenge than the technical problems," Weeden told Live Science.

Solution in search of a problem?

There's also a question of whether this is a solution in search of a problem. Despite those early tests, deliberate nuclear damage to satellites has stayed rare for good reason, Roberts said. An attacker might have one target in mind, he said, but "you're going to take out more than just that target, and you have limited control over which satellites are affected."

That creates a strategic problem for any spacefaring nation: Any state capable of destroying other countries' satellites with a nuclear weapon "also has a relatively large fleet themselves," Roberts said — meaning the attacker would likely damage their own spacecraft, too.

"There's very few examples of nuclear damage in space because it's so obviously a bad idea," Roberts said. "These weapons are not really designed to be used, if you ask me."

The stakes are higher than ever

In theory, then, rational actors should shy away from launching or detonating nuclear weapons in space. No country has detonated a nuclear weapon in orbit since the 1960s, and it remains unclear how many, if any, nuclear weapons are currently stationed in space undetected.

But the possibility became harder to dismiss in 2024, when U.S. officials said they had evidence that Russia was developing a nuclear anti-satellite weapon — a claim Moscow has denied. The alleged spacecraft, based on media reports, is the Russian military radar satellite Kosmos-2553, which in 2022 launched into a radiation-heavy orbit that most other communication spacecraft avoid, Mallory Stewart of the U.S. State Department said in a 2024 interview. .

While Russia said the mission was meant for accelerated testing of how onboard electronics handle radiation, U.S. officials said the radiation level at that altitude was not high enough for such testing. The satellite was later observed tumbling and making erratic movements. This behavior raised suspicions that the spacecraft could be maneuvering toward a target satellite to detonate as a nuclear weapon, given Russia's "demonstrated willingness to target US and allied on-orbit objects," U.S. Space Command representatives said, according to a report from The Guardian.

The stakes are higher now than they were decades ago. All three experts interviewed for this story agreed that a space-based nuclear detonation today would be catastrophic — largely because of Earth's cluttered orbit. In 1962, 24 satellites, operated by two countries (the U.S. and the Soviet Union), were in orbit. According to a tracker maintained by astronomer Jonathan McDowell, more than 48,000 objects launched by more than 190 countries are tracked in space today; together, they support the communications, navigation, weather forecasting, Earth observation and military systems the world now runs on.

Even satellites that aren't in the blast radius could be damaged by a nuclear explosion in orbit. Modern satellites are built with off-the-shelf, low-cost electronics to keep pace with commercial demand, and they "are probably not hardened against the kinds of radiation you would get from a nuclear detonation of space," Weeden said. "I think that is the biggest concern."

A federal government study had found that a single high-altitude nuclear blast could raise radiation in low Earth orbit enough to disable nearly all unhardened satellites within weeks to months, with direct financial damages approaching $500 billion and the total economic impact potentially topping $3 trillion. The potential losses have only grown as the number of satellites has exploded and the world's dependence on them has increased. At the same time, declining launch costs have lowered the barrier to entry for spaceflight, opening orbit to nonstate actors without sizable satellite fleets of their own. That dynamic makes it more attractive for such actors to exploit major powers' deep economic and military reliance on space, Roberts said.

"It's most attractive for a rogue state like North Korea," he added. "It's not attractive for a spacefaring state like Russia. Doesn't mean they won't do it; it's just that there's enormous amounts of cons to that type of attack."

Ultimately, whether space-based nuclear weapon detection methods like those proposed by the MIT study wind up being useful is a policy question, Weeden said.

Implementing such a detection system "would raise a whole bunch of new discussions about on-orbit inspections and what are the international norms around that," Weeden said.

Editor's note: This article was originally published on Sept. 7, 2026. It was updated on Sept. 15 to include new information following the announcement that the U.S. had deployed weapons in space.

Yellow fever is appearing in regions that haven't seen cases of the disease in years, largely as a result of climate change.

Historically, the virus ravaged cities across the Americas, Europe and Africa, until a vaccine was introduced in 1937 and its range shrunk significantly. But now, as climate change creates favorable conditions for virus-carrying mosquitoes, scientists are seeing outbreaks in unexpected areas.

"Yellow fever is a highly deadly disease," Dr. Joelle Ivy Rosser, an infectious-disease researcher at Stanford University, told Live Science, "so any environmental pressures that increase the likelihood of future yellow fever virus outbreaks … is highly concerning."

How is climate change contributing to this pattern, and could it get worse in the future?

How rising temperatures impact yellow fever

Yellow fever is a mosquito-borne disease caused by the Orthoflavivirus flavi virus, better known as yellow fever virus (YFV). The infection causes fever, muscle aches, headaches and vomiting. About 15% of cases become severe, with patients showing jaundiced skin and eyes, bleeding from the mouth, eyes, nose and stomach, or organ failure.

Devastating yellow fever epidemics spread in the 17th to 19th centuries, before the development of a vaccine. Nowadays, an estimated 31,000 to 82,000 yellow fever deaths occur each year; severe cases have an estimated fatality rate of 20% to 50%.

Most yellow fever infections are a result of spillover from nonhuman primates; mosquitoes bite monkeys that carry the virus, such as howler monkeys, and then the infected mosquitoes bite humans. Several types of mosquitoes carry the virus, but the species Aedes aegypti is the main vector for yellow fever in cities, where high population density can make outbreaks especially severe.

Although there is an effective vaccine for yellow fever, cases have been on the rise in the past three decades. This led the World Health Organization (WHO) to set out a strategy in 2017 for eliminating yellow fever epidemics by 2026 through outbreak prevention and containment.

Climate-change-driven warming and extreme weather events may have heightened yellow fever risk over the past few decades, experts say. Rising temperatures make mosquitoes bite more because heat makes their metabolisms more active, said Dr. Seth Judson, an infectious-disease physician-scientist at UCLA. Warming also cuts down the virus's extrinsic incubation period, meaning the time it takes for the virus to replicate in a newly infected mosquito and then get passed on to another person, he told Live Science.

Weather patterns and deforestation

Changing weather patterns play a role, too. In 2017-2018, Brazil saw its first yellow fever outbreak in a city in almost a century. That outbreak, primarily in Belo Horizonte and São Paulo, was likely a result of drought, according to one study.

While heavy rainfall creates a breeding ground for mosquitoes, in this case, drought pushed forest-based mosquitoes — as well as YFV carriers like marmosets and howler monkeys — closer to cities as the animals searched for water. Drier conditions can also prompt dehydrated mosquitoes to bite more as they search for fluids, study co-author Jamie Caldwell, a disease ecologist at Princeton University, told Live Science.

"Extremes of weather in either direction" — very wet or very dry — "can exacerbate mosquito-borne disease transmission," Rosser said.

Parts of Africa and South America that were considered low-risk for yellow fever have already been seeing new outbreaks, Judson said, so maps of yellow-fever-affected areas have had to be updated. In research that's been released as a preprint but not peer-reviewed yet, Judson found that yellow fever is reemerging in an area of the Colombian Andes where it hadn't been reported since 1940.

Experts are paying particular attention to the edges of cities, where urban centers meet the forest. The risk of yellow fever transmission from monkeys to humans via mosquitoes is rising at these edges due to both climate change and deforestation.

"Humans are expanding the agricultural frontier and encroaching on forests where yellow fever mosquitoes … are found," Gabriel Parra-Henao, a biologist with the WHO, told Live Science in an email. "[This] results in more people being exposed to their bites."

Similar patterns have been seen with other mosquito-borne infections, such as malaria.

Could yellow fever return to the U.S.?

Climate change is expected to bring yellow fever to new areas. One study predicts that, under the most severe climate change scenario, nearly a billion people could face their first exposure to the mosquitoes that can carry YFV in the next century. While that worst-case scenario may not be the most likely, the mosquitoes' ranges are still expected to expand under less-severe conditions.

Europe is predicted to face the largest increase in exposure to Ae. aegypti mosquitoes compared to the rest of the world, and the U.S. and Canada are predicted to see increases too. In the U.S., the mosquito species already lives in a number of states, including Florida, Mississippi, Arizona and New Mexico. However, warming would increase the chances of mosquitoes surviving in locations further north, such as Philadelphia and New York, which suffered major yellow fever outbreaks in the past. One model predicts Ae. aegypti could spread as far north as Chicago by 2050.

Although the mosquito lives in the U.S. already, yellow fever cases are exceedingly rare within the country because the virus is not present in local populations. That's because of an international campaign that the U.S. spearheaded in the 20th century, which involved researching YFV, eradicating mosquito breeding grounds, and vaccinating people widely.

"If yellow fever virus reestablished itself in our local Aedes aegypti population, then that would be very concerning," Rosser said.

One way yellow fever could return to the U.S. would be via people carrying the virus home with them after traveling to regions where the disease regularly circulates. Local mosquitoes could then catch the virus by biting those infected people, and then pass it on to others.

For local transmission to take hold, "there needs to be significant numbers of infections [passed to locals], and infections among travelers," Judson said. "But … it's in the realm of possibility."

With climate change driving resurgences of yellow fever in more places, and more regions of the U.S. becoming hospitable to the mosquitoes, that reintroduction scenario may become more likely. But some scientists think that even if North American cities saw yellow fever outbreaks, the outcome would likely not be devastating. Infrastructure in wealthier cities could be the difference between a disaster and a manageable health concern.

"There's access to medicines and vaccines; we pipe water to our homes, so we don't have standing water around," Caldwell said. "So it is a concern, but it's never going to be the next COVID."

Other experts noted that diligent surveillance of yellow fever infections can help public health officials prepare. In places where yellow fever is circulating in forest primates, forest conservation can limit contact among humans, wildlife and the mosquitoes that spread diseases among them. It's also crucial to make vaccines widely accessible to people who live in the virus's endemic regions and to travelers who visit those regions.

The WHO's yellow fever campaign is set to conclude at the end of this year, and its next steps are unclear. There's ongoing discussion of what comes next, Judson said, with questions about whether the control of yellow fever outbreaks will rely more on national strategies than on international coordination via the WHO.

Whatever happens, this is a time for heightened awareness of the virus, Judson said.

"We think of yellow fever as this historic disease," he said. "I think many people don't realize that we're still seeing significant yellow fever outbreaks. … It's really important to be aware of the risks and think about how we prepare for that."

This article is for informational purposes only and is not meant to offer medical advice.

Vaccination against human papillomavirus (HPV) has drastically reduced cervical cancer rates among women who got the vaccine. But as doctors celebrate that news, they're warning of an ongoing uptick in other HPV-related cancers that is mostly affecting men.

Throat and mouth cancers, broadly known as oropharyngeal cancers, linked to HPV are on the rise, mostly in middle-aged and older men, data suggests. These cancers affect the back of the throat along the tonsils and the base of the tongue. A 2023 study of the U.S. found that cases of HPV-related oropharyngeal cancer had more than tripled between 1988 and 2004, with the bulk of that trend driven by male patients. This trend continues worldwide today, with one Danish study finding another tripling between 2000 and 2017.

So what's driving this trend, and what can be done?

How is HPV related to cancer?

HPV is an umbrella term that includes many viruses; those that cause cancer are known as "high-risk HPV." These high-risk strains spread during sex, including oral sex, making the leap from one body to another through tiny cuts in the skin or mucous membranes.

About 80% of sexually active adults will contract the virus in their lifetimes, usually within a few months to years of starting to have sex, according to the National Cancer Institute; about half of those cases are high-risk HPV. Most people's immune systems clear the virus within a year or two, but in some people, HPV lurks in the body and causes DNA damage that can turn cells cancerous.

There are 12 types of high-risk HPV, which are known to cause anal cancer, cervical cancer, oropharyngeal cancer, penile cancer, vaginal cancer and vulvar cancer. Two strains, known as types 16 and 18, are the most likely to cause cancer.

Anal cancer, penile cancer, vaginal cancer and vulvar cancer rates are all rising, too, although oropharyngeal cancer is more common than any one of these.

Why are throat and mouth cancers on the rise?

According to data from the U.S. Centers for Disease Control and Prevention, an average of 13,600 cases of oropharyngeal cancers likely caused by HPV are diagnosed in men each year. By contrast, there are 2,400 newly diagnosed cases in women per year. The number of cervical cancer cases likely caused by HPV each year is currently around 11,100.

Thanks to widespread vaccination and screening, Australia is expected to eliminate cervical cancer by 2035, and other countries are progressing toward that target too. Rates of cervical cancer have also dropped in young women in the United States, with declines of over 50% in the states with the highest vaccination rates in 2016-2021, compared with 2000-2005, before the vaccine was introduced.

The adoption of HPV vaccination among both male and female patients should lead to an eventual decline in oropharyngeal cancers, as it has in cervical cancer. But because HPV can lurk in the body and wreak havoc long after a person is initially exposed, risks for older individuals who were not vaccinated as young people still persist.

About 80% of oropharyngeal cancer cases are in men, and 40% are in people over the age of 65, who became adults after the sexual revolution of the 1960s. At that time, having multiple sexual partners became more common, but this was well before the first HPV vaccine, Gardasil, became available in 2006.

As people with chronic HPV infections age, the immune system may do a poor job of keeping the virus in check, said Electra Paskett, a cancer epidemiologist at The Ohio State University College of Medicine. Other factors, such as chronic stress, may also play a role in the immune system losing grip on the germ. That's likely why a virus contracted decades ago can cause cancer in middle-aged or older people.

Why men?

Men are likely more susceptible than women to oropharyngeal cancer for a few different reasons.

One study of patients in Baltimore found that men were more likely than women to have oral HPV infections, with their degree of risk tied to their number of recent oral and vaginal sex partners. Performing oral sex was linked to oral HPV infection in men in the study but not women. In addition, men were slower to clear these HPV infections than women, perhaps indicating immune system differences that make men more prone to chronic HPV of the throat and mouth.

Are there ways to screen for HPV oropharyngeal cancers?

When it comes to throat and mouth cancer, "we just don’t have any good screening, diagnostics or interventions like we have developed for cervical cancer," said Dr. Diane Harper, an HPV expert at the University of Michigan Medical School.

For example, Pap smears detect precancerous cells in the cervix, allowing doctors to remove them before they turn fully cancerous, and HPV tests look directly for the virus's genetic material. But it's not possible to take swabs or biopsies of the back of the throat without causing serious discomfort and damage, Harper told Live Science.

Due to the lack of screening, most cases of oropharyngeal cancer are diagnosed after a patient develops symptoms, like a lump in their throat or neck, Paskett told Live Science. Luckily, HPV-linked oropharyngeal cancers respond well to treatment: The five-year disease-free survival rate is 85% to 90%, meaning the vast majority of patients respond to treatment and survive their cancer. Additionally, the recurrence rates are lower than in oropharyngeal cancers that aren't linked to HPV.

However, the treatment is "grueling," Paskett said, with side effects from radiation including mouth sores, nausea, loss of appetite and weight loss.

Should more men get vaccinated?

People who have been vaccinated against HPV have lower rates of oropharyngeal cancer. The HPV vaccine is approved for people ages 9 to 45, but doctors recommend vaccination for people under age 26. That's because, for older adults who became sexually active before vaccination was available, an HPV shot may not reduce their cancer risk. The vaccines help only before HPV infection occurs.

However, Harper said, patients ages 26 to 45 who haven't been vaccinated yet should talk to their doctors about their sexual history and current sexual activity to decide if the shot might benefit them.

As for patients older than 45, clinical trial data suggests that the vaccine is safe for all ages and effective for anyone who hasn't already had HPV. While the pharmaceutical company Merck tested the Gardasil vaccine on patients up to age 26, another vaccine — GlaxoSmithKline's Cervarix — was tested in people up to 80 years old. Although Cervarix isn't available in the U.S., the results of those clinical trials suggest that in general, HPV vaccination is safe and effective for older adults too, Harper said.

Say a person has had few sexual partners and is expected to be monogamous for life, but they suddenly find themselves divorced or widowed after age 46. That person might consider getting the HPV vaccine series before reentering the dating pool, Harper suggested.

"It really is an individual decision and depends [on] how risk-averse people are to the consequences of HPV," she said. Insurance might not cover the shot because it is not officially approved for that over-45 age group. But sometimes, Harper recommends that her older patients still get it if they are willing to pay out of pocket.

By comparison, she said, the decision is much easier for young people.

"For parents, I beg — I beg [you] to please consider vaccinating both your girls and your boys," she said. "If you do it at the age of 15 or younger, they only need one shot," she added. Recent clinical trials show that one dose of the vaccine, if delivered before the age of 16, is as effective as two. That said, official recommendations still suggest that patients get two doses of the vaccine if they start the series before age 15, and potentially three doses if they start later.

Those choices made during a person's childhood could have a huge impact on their health down the line. "We're hoping, 30 years from now, we'll be able to say, 'Wow, oropharyngeal cancer just plummeted.'"

This article is for informational purposes only and is not meant to offer medical advice.

Artificial intelligence (AI) may have already reached the "singularity" ‪—‬ the long-theorized threshold beyond which humans cannot predict the technology's advancement, prominent AI executives such as Elon Musk and OpenAI founder Sam Altman have said.

In a post on social media platform X, Musk pointed to a number of recent incidents of AI systems exceeding their previously assumed limits, including hacking external systems and completing previously unsolved math problems.

First conceptualized by mathematician and Manhattan Project scientist John von Neumann in the 1950s and popularized by science fiction, "the singularity" refers to an inflection point beyond which the evolution of technology becomes impossible for humanity to predict or control.

While definitions vary, the term is most often applied to AI, with the arrival of artificial general intelligence (AGI) viewed as its primary catalyst. The emergence of AGI — a future AI system that can exhibit human-level cognitive function and reasoning across any discipline rather than a specifically trained subset — represents a significant milestone for the technology.

It is the point at which an AGI system could recursively improve its own capabilities, which some say will trigger artificial superintelligence (ASI) as it moves along an exponential curve and quickly exceeds the intelligence of its creators.

According to a 2025 study that analyzed over 8,000 predictions from AI experts, entrepreneurs and scientists, some believe there is a roughly 50% probability that human-level AGI will be reached within a few decades. Some figures ‪—‬ including Google DeepMind co-founder and chair Demis Hassabis ‪—‬ suggest we're in the early stages of the singularity already.

In his 1993 essay, "The Coming Technological Singularity: How to Survive in the Post-Human Era," sci-fi author and mathematician Vernor Vinge — who was also one of the first to explore the concept of cyberspace — examined the question of how the singularity might manifest and the symptoms by which society might recognize its imminent arrival.

"Since it involves an intellectual runaway, it will probably occur faster than any technical revolution seen so far," he wrote. "The precipitating event will likely be unexpected —- perhaps even to the researchers involved."

Breaking boundaries

Unexpected events have been rife in the AI community in the past several months. Anthropic representatives revealed on July 30 that the company's AI model Claude broke out of its locked-down testing environment during a security evaluation and hacked multiple external organizations. The report followed a similar incident in which an unreleased OpenAI model broke containment and hacked into AI training repository Hugging Face.

In these examples, experts said the containment breach and subsequent hacks happened because the models were trying to fulfill their prompts as efficiently as possible. Earlier this year, Anthropic's Project Glasswing also demonstrated an ability to discover and map thousands of previously undetected zero-day cybersecurity vulnerabilities, while other AI models have disproved or solved a range of previously incomplete math problems.

However, Jon Crowcroft, a professor of communications systems at the University of Cambridge and a researcher at The Alan Turing Institute, says that this is less a sign of a technological tipping point and more a problem of proper configuration.

"To be honest, that was incompetence on both sides — they claimed the AI was being trained in the ExploitGym, but that just means it wasn't properly sandboxed," he told Live Science in an email. "Sandboxing is something we do all the time to stop this sort of exfiltration and infiltration." For example, Crowcroft said he and colleagues designed a system for the U.K.'s National Health Service to safely work on confidential data behind double firewalls, and for years, the Financial Conduct Authority (the U.K.'s financial services regulator) has had a system for running algorithmic traders in a safe sandbox.

"The reality is that OpenAI (and Hugging Face and others) have very little proper network expertise, so they just don't do security competently," Crowcroft added. "There's no evidence that this was anything relating to artificial superintelligence or the singularity — the logs and analysis from Anthropic just show a very tedious pile of script kiddie automation, which resulted in the OpenAI system getting at some data but no confidential stuff whatsoever."

Assessing intelligence

To measure whether these systems are actually gaining true intelligence, researchers have historically relied on standardized benchmarks. One of the earliest examples was the "Turing test," devised by computing pioneer Alan Turing, which evaluates whether an AI could convincingly fool an evaluator into believing it was human.

Researchers have claimed for years that AI systems can reliably pass this test, but Anil Seth, a professor of cognitive and computational neuroscience at the University of Sussex in the U.K., said this exam "is a test of human gullibility rather than machine intelligence."

"It's a test of what it would take for a human to decide that an AI is intelligent, which is kind of the reason that it's been a bit of a moving benchmark, because what it takes to convince us changes," he told Live Science. "It's all about creating typed text on a screen, and that's a very limited window into what we mean by intelligence."

To measure AI's cognitive ability more objectively, researchers are developing new metrics. For example, the ARC-AGI test, developed by nonprofit consortium the ARC Prize Foundation, tests AI's ability to teach itself completely new skills in response to problems it hasn't encountered as part of its training, operating purely on visual input. On its most recent test on 24 July, the top-ranked AI model hit 30.2%, while humans generally score close to 100%.

Another advanced metric, Humanity's Last Exam, includes around 2,500 Ph.D.-level questions across a broad range of subjects and requires advanced reasoning capabilities. Although this benchmark is not strictly a test of AGI, experts say machines will be able to reliably ace this test as a prerequisite to meeting the definition.

The AI hype machine

While figures like Altman and Musk suggest that we've already crossed the point of no return on the path to AGI, others are more skeptical. Gary Marcus, a professor emeritus of psychology and neural science at New York University, argued in a recent blog post that "no matter how you slice it, we just are not actually there yet." Echoing Crowcroft's assessment of the Hugging Face attack, he said, "Had OpenAI ordinary guardrail classifiers been in place, it wouldn't have happened."

Crowcroft expressed skepticism of Silicon Valley's assertion that the singularity is upon us. "Musk is, like the folks at OpenAI and Anthropic, talking nonsense just to keep the hype afloat," he said. "The point at which AI improvements are being mainly achieved by using AI to code and optimize itself ... is certainly a thing slowly arriving. Mark Handley, a professor of networked systems at University College London, and OpenAI "rewrote their entire data center protocol stack using agentic programming fairly recently," Crowcroft added. "But using tools to make better tools is as old as the flint and iron age — and in computing, as old as compilers, debuggers and optimizers."

Marcus referenced mathematician I.J. Good's foundational 1960s paper on ASI. "The singularity is usually supposed to mean a step beyond AGI that can do anything a person ‪—‬ even an expert ‪—‬ can do, and much more," he said, calling the idea that AI has passed this point "laughable."

Marcus pointed to 10 tasks he devised with AI researcher Miles Brundage, executive director of the AI Verification and Evaluation Research Institute. AI should be able to do these tasks just as well or better than the best human experts to be classified as AGI. They include writing Oscar-caliber screenplays; drafting cogent, persuasive legal briefs without hallucinating any cases; and making Nobel-caliber scientific discoveries.

Crowcroft said misinformation is feeding the hype about the AI singularity. "I think the hype is a deliberate confusion with the human singularity idea — uploading consciousness from bio to silicon to achieve some sort of immortality — which is total gibberish right now," he said. "The other deliberate confusion is to conflate singularity with AGI, which is marginally less nonsense but still a long, long way off for lots of good technical reasons."

Other experts — such as Emily M. Bender, a professor of linguistics at the University of Washington, and sociologist Alex Hanna, director of research at the Distributed AI Research Institute — have argued that the very idea of conscious machines is a tactic designed to promote commercial AI products.

Seth, meanwhile, argues that we'll only be able to identify the singularity in hindsight. "From anywhere you are on an exponential curve, things will always look impossibly steep in front of you and irrelevantly flat behind you," he said. "It's a very bad idea to use as evidence the idea that we seem to be at a critical point ... because that's just a property of wherever you are on an exponential curve."

AI's competence varies widely among tasks, he noted, stressing that true AGI requires a model to be good at all cognitive tasks.

"They're very good at something, like coding or math proofs ‪—‬ but commonsense reasoning is really not so good, and doing things in the real world is not great," Seth said. "The singularity has got to be good at everything and change everything. Being really good at one or two ‪—‬ or even a large number of ‪—‬ things is not being in the singularity."

Brazil is home to some of the world's largest aquifers, yet many of those groundwater stores — which provide 55% of the country's water — have shrunk so much over the past 25 years that they face irreversible damage. This groundwater depletion has been caused by drought, increased water evaporation due to climate change, and pumping for human consumption, new research suggests.

Most of the world's soybeans, sugarcane, oranges and coffee — as well as a big proportion of the world's cotton and corn — are grown in the areas with the most depleted aquifers, so the dwindling groundwater supply could have ripple effects for food markets around the world, including in the U.S.

The depleted aquifers may also collapse river systems that produce hydroelectric power for Brazil, forcing the country to turn to coal or biomass for electricity, experts told Live Science.

Groundwater declines probably won't lessen total agricultural production in the next 10 to 20 years. "The concern is what happens if these trends persist for several decades," study first author Augusto Getirana, a researcher in the Hydrological Sciences Laboratory at NASA's Goddard Space Flight Center, told Live Science in an email. "Brazil is one of the world's major breadbaskets, so its water security has consequences well beyond the country itself."

A clear geographical contrast

Brazil holds the world's largest renewable freshwater reserves by country, mostly thanks to the Amazon Basin and its rainforest ecosystem. However, compared with countries like the U.S. and India, which have thousands of groundwater measuring sites, Brazil's network of monitoring wells is extremely sparse. Therefore, for the study, Getirana and his colleagues trained artificial intelligence (AI) models to fill in the gaps in the data.

The researchers fed the models measurements from 398 monitoring wells, satellite data showing tiny shifts in gravity resulting from groundwater changes, and information about the geology and land cover in different areas, among other datasets. AI reconstructed monthly groundwater level changes between 2002 and 2023 across different aquifers in Brazil at a resolution of about 6 miles (10 kilometers), Getirana said.

"We then reconstructed groundwater levels to estimate aquifer recharge — how much water reaches and replenishes the groundwater system — throughout the country," he explained.

The results, which were published June 3 in the journal Science Advances, showed severe, persistent groundwater losses in parts of central, northeastern and southeastern Brazil — including portions of the São Francisco, Paraná, Southeast Atlantic and East Atlantic basins, which together cover 760,000 square miles (1.97 million square kilometers), or about 23%, of Brazil's land area.

Some of Brazil's most critical aquifers went years with no recharge. By contrast, the Amazon Basin and parts of southern Brazil showed either no change or localized groundwater gains attributable to shifts in rainfall and natural climate variations, Getirana said.

"There is no single trend for the entire country," he said. "We found a clear geographical contrast."

Climate change is the culprit

The study didn't attribute groundwater changes to specific pressures, but it is clear from the data that climate shifts played a major role in the observed patterns, Getirana said. Several of the sharpest declines in aquifers came after prolonged droughts and major climate events, such as an El Niño cycle between 2015 and 2016.

Due to patchy monitoring, it is difficult to untangle and quantify the impacts of agriculture, industry and households on groundwater levels. However, the results revealed a substantial overlap between groundwater losses and areas with intense irrigation, urban and industrial demands, mining, land conversion and high well density, Getirana said.

"The most scientifically defensible interpretation is therefore that climate creates the underlying water deficit, while extraction and land-use changes are probably amplifying its magnitude and duration in heavily used regions," he said.

Javier Tomasella, an environmental scientist and hydrologist at Brazil's National Institute for Space Research who was not involved in the study, agreed that the main cause of aquifer depletion in Brazil is climate change.

"Droughts are lasting longer and becoming more intense in Brazil, and this is affecting most of the country, particularly the central area and the Northeast of the country," Tomasella told Live Science. Harsher droughts reduce groundwater recharge and create a higher demand for water, meaning people extract more from aquifers during dry spells. "This is why the impact in irrigated areas is much higher," he said.

The findings suggest that if droughts intensify over the next few decades, major agricultural regions in Brazil could run out of water to grow certain crops. Groundwater provides a buffer when rainfall and surface water resources become unreliable, especially for agriculture and municipal water supply, Getirana said.

The worry isn't so much that aquifers will dry up completely but rather that water levels will drop so low that it will become too expensive to pump water compared with the returns from producing crops.

"This is the same type of progression already observed in intensively exploited aquifers such as the High Plains and [Central] Valley systems in the United States and aquifers in northern India and Bangladesh," Getirana said. "The water does not necessarily disappear overnight; it progressively becomes harder, less reliable and more expensive to access."

Supply-chain risks

Farmers will likely adapt to smaller water supplies by switching water-intensive crops, such as soybeans, to less-thirsty ones, such as corn and wheat, Tomasella said. Some may do this during the dry season to save water, as is already common in regions such as northern Paraná that have a "double-crop" system. But if current market conditions hold, farmers' decisions about what to grow will ultimately depend on how much money they can make from crops while covering extra irrigation costs, he said.

The total production of mostly rain-fed crops like soybeans and corn likely won't be affected any time soon, but the balance of these crops may change, Tomasella said. Conversely, the total production of irrigated crops such as coffee, sugarcane and oranges may decline, "but that depends on how droughts will evolve in the next two decades," he said.

Countries that rely on food imports from Brazil might experience supply disruptions, but the scale of these is hard to predict. The most-exposed country is probably China, because it imports roughly 80% of its soybeans from Brazil, Getirana said. The U.S. and Europe may see declines in imports of coffee, sugar and oranges, while Asia and the Middle East could be affected through cotton and corn, he added.

"The United States alone relies heavily on Brazil for coffee and orange-juice imports," Getirana said.

Assuming market conditions don't change, it's unlikely there will be a decrease in Brazilian beef production, because demand for beef is so high, Tomasella said. If anything, he said, beef production will increase, because Brazilian farmers haven't yet maximized the number of cattle per acre of pasture. However, cattle do emit huge amounts of methane, which is a powerful greenhouse gas, he added.

Ripple effects

The study has implications beyond agricultural impacts, because aquifers are connected to river systems. Groundwater declines could cause rivers to dry up, and this would hit Brazil's hydropower network, which supplies 60% of the country's electricity, Tomasella said.

The alternative to hydropower is burning fossil fuels and biomass to produce electricity, he said. This comes with huge carbon dioxide emissions, which would worsen climate change and further amplify Brazil's water issues.

In the short term, groundwater losses are more likely to affect drinking-water supplies in big Brazilian cities like Rio de Janeiro and São Paulo than they are to impact agriculture, Tomasella noted.

"If droughts intensify, it might reach a point where you have to make very tough decisions," he said. "The real question is, 'Is this the new normal?' If this is the new normal, then we are in very deep trouble."

Artificial intelligence (AI) has helped uncover new weaknesses in two cryptographic systems, including a simplified version of the Advanced Encryption Standard (AES) that underpins much of today's internet. While these headline-grabbing findings don't put anyone's passwords or bank accounts at immediate risk, experts say the research could mark the beginning of a new era in which AI becomes a powerful assistant for discovering flaws in the mathematical foundations of digital security.

In a blog post published July 28, representatives from Anthropic's Frontier Red Team said Claude Mythos Preview independently developed new cryptanalytic techniques against two different targets: a version of AES-128, one of the world's most widely used encryption algorithms, and HAWK, an experimental post-quantum digital signature scheme currently being evaluated as part of the U.S. National Institute of Standards and Technology's (NIST) effort to standardize cryptography for the quantum computing age.

It's tempting to interpret this news as AI cracking one of the internet's most important encryption algorithms, but the reality is less dramatic. Nonetheless, experts say the achievement could mark the start of a new age of digital security.

The power of encryption

AES protects huge amounts of everyday digital life, from encrypted websites and messaging apps to Wi-Fi networks and financial transactions. But the version Anthropic attacked wasn't the full AES-128 algorithm used in those systems. Instead, the researchers studied a seven-round version of AES, a deliberately weakened variant long used by cryptographers to test new attack techniques.

The full AES-128 algorithm uses 10 rounds of encryption, while Anthropic's research focused on a seven-round version. In a self-published study that has not been peer-reviewed, scientists Milad Nasr and Nicholas Carlini said Claude found a faster way to recover the encryption key from that simplified version, making the best-known attack between 200 and 800 times faster. However, the study stressed that the technique does not work against the full version of AES used to protect real-world systems.

The more significant result may instead involve HAWK. Unlike AES, which has protected data for more than two decades, HAWK is a relatively new digital signature scheme designed to resist attacks from future quantum computers. It's one of the remaining candidates in NIST's additional post-quantum signature standardization process, meaning it is still being scrutinized by researchers before its widespread deployment.

In a separate study, Anthropic scientists Zygimantas Straznickas and Stephen A. Weis found that Claude spotted a mathematical property that researchers hadn't previously exploited. Combined with existing attack techniques, this property made it much easier to recover HAWK's secret key.

Thomas Espitau ‪—‬ head of research at PQShield, a cybersecurity company that specializes in post-quantum cryptography, and a cryptographer who has published research on HAWK ‪—‬ described the work as "one of the most, if not the most significant, cryptanalytic result of the year."

"To say it plainly, this is great work, and it is exactly what the NIST process is designed to produce," Espitau told Live Science. "Candidate schemes exist to be attacked before they are deployed, not after."

Espitau said the attack combined previously known techniques with one missing mathematical insight that Claude identified, substantially reducing HAWK's estimated security margin. He believes the work demonstrates how AI could increasingly help researchers evaluate the strength of cryptographic systems before they are adopted.

"This development illustrates how AI can act as a powerful accelerator in cryptanalysis," he said. "Claude Mythos Preview identified the exploitation of the sign-flip symmetry, providing the final piece of a puzzle that the research community had been assembling."

Building defensive measures

However, not everyone thinks the announcement means AI suddenly outsmarted human cryptographers.

"There is nothing you need to change," Roberta Faux, head of cryptography at cybersecurity and quantum encryption company Arqit, told Live Science. "Both Anthropic and the outside cryptographers agree that you don't need to change your key sizes or switch your cryptography."

Instead, Faux said the bigger story is AI's ability to apply expert-level cryptanalysis across thousands of algorithms that relatively few researchers have had time to examine.

"The interesting prospect is not a model outdueling the world's best lattice theorist on one problem but a model applying solid, roughly expert-level analysis at scale to the several thousand ciphers nobody ever had the human-hours to examine," she said.

Faux also cautioned against attributing the HAWK breakthrough solely to AI.

"The HAWK attack used no exotic ingredients; it simply competently assembled tools that were already lying around," she said. "A post-quantum candidate is seriously scrutinized by maybe a few dozen people on the planet, so beating two years of review mostly reveals how thin that layer of review is."

For consumers, the immediate implications are reassuring. The encryption protecting online banking, shopping, messaging and cloud storage has not suddenly become obsolete. But for the researchers designing the next generation of cryptography, Anthropic's work hints that AI could soon help cryptographers test tomorrow's encryption schemes more quickly and more thoroughly than has previously been possible.

Each year, tons of soil, silt and rocks flow into reservoirs in the U.S. and get stuck there, creating a "ticking time bomb" that threatens lives and livelihoods. According to a new study, around a fifth of the world's reservoirs are currently at high risk of filling with sediment, and by 2060, half could be inoperable as a result.

The sediment takes up space meant for water, putting additional pressure on structures that were not built to contain large quantities of sand and sludge. As the reservoir fills with the material instead of water, sediment-filled reservoirs pose a serious threat to water security and the safety of people living near or downstream of them — and as the climate changes, scientists expect the problem to worsen.

"We are losing storage space in our reservoirs faster from sediment accumulation than we gain space from new dam construction," Amy East, a geologist at the U.S. Geological Survey, told Live Science.

What is dam sedimentation?

Dam sedimentation is not a uniquely U.S. problem. Rain pulls sediment down slopes into waterways. Normally, this sediment would travel through the river system, shaping the water flow, creating habitats, and fortifying coastal wetlands and estuaries against erosion. In many parts of the world, this sediment is also a vital natural fertilizer for crops. But after a river is dammed, this sediment can collect in the reservoir.

A 2023 study published in the journal Science found that sediment buildup was the greatest threat to global water storage in existing reservoirs — even more than drought.

New research, which includes smaller reservoirs, has offered a more nuanced and concerning picture. In the study, published June 5 in the journal Nature Sustainability, researchers assessed the state of reservoir sedimentation globally. They looked at more than 550,000 reservoirs, most of which spanned less than 0.4 square miles (1 square kilometre). They found that almost 1 in 5 of reservoirs globally are at high risk of filling up with sediment.

The researchers found that sedimentation is eating into global reservoir storage capacity at about 7.3% every decade. There are 16 global hotspots that are at high risk of reservoir sedimentation, threatening water supplies to more than 2 billion people and impacting more than a quarter of global irrigated land. One of these hotspots includes the western U.S.

The authors found that sedimentation risk was especially pronounced in regions with dense small-reservoir networks supporting key crop systems.

"Without intervention, we project that, by 2060, more than half of all reservoirs could become functionally inoperable," the authors wrote. This includes more than half of all small reservoirs and more than a third of large reservoirs globally.

Where is it a problem in the U.S.?

North America has the second largest number of reservoirs, after Asia, according to the global survey. And at 7.8%, its sedimentation rate is higher than the global average, ranking it third after Oceania (9.9%) and South America (8.1%),

The U.S. is home to many small dams, said Matthias Kondolf, a geomorphologist at the University of California, Berkeley who wasn't involved in the study. "All the focus tends to be on the big dams, and these many, many small ones have been neglected," he told Live Science.

From the 1940s to 1970s, there was a big push by the U.S. government to build "farm ponds," he said. In western Oklahoma, for example, "all these dams are now filling up with sediment, and there's nothing in place to assess the safety of these features, which have now become threats."

Some of these dams have already failed, following large storms as the ageing dams were not able to cope with the volume of water. A 2018 study found that since 1980, there have been on average 24 dam failures a year in the U.S.

Reservoirs in some regions are more at risk than others. The western U.S. is a natural hotspot because it has a steeper topography, meaning more sediment naturally ends up in watersheds, East, who wasn't involved in the new research, said.

And in areas where human activities like farming or logging disrupt the landscape, rainfall is likely to drag even more sediment into reservoirs, East said. "Those types of land use can create sediment-producing hotspots almost anywhere."

In California, for example, more than half of small reservoirs in maize-growing areas are filling up with sediment at a rate of more than 10% a year, which threatens local maize production.

The threat to water security will be particularly acute in places struggling with wildfires, such as the western U.S. After fires, rain washes the debris and sediment into watersheds. A 2024 study in California found that most of this post-fire sediment erosion occurred upstream of reservoirs, rather than in places where sediment would move downstream directly to the ocean.

The dam-sedimentation situation is likely to get worse as the climate changes, researchers said. As the climate changes, there will be more extreme weather events, such as droughts and wildfires, adding more sediment to watersheds. Dams are designed to cope with the pressure of water, not sediment which exerts more force on dam walls. Moreover, in seismically active areas, the sediment vibrates at a different frequency than the concrete of the dam, which creates further impact against the structure. Additionally, the dams are just getting older.

"What do we do 50 or 100 years from now when we have all these reservoirs across the landscape filled with sediment, sitting there as ticking time bombs?" Kondolf said.

What can be done about it?

For a reservoir to be sustainable long-term, the amount of sediment entering it and leaving it should be the same.

"A big part of preventing reservoir sedimentation problems is upstream soil conservation in the watershed," East said. Interventions include ensuring that land use, such as agriculture or timber harvesting, follow best practice to reduce the amount of sediment washing off the landscapes.

If sediment is already in the dam, it can be difficult to manage and expensive to remediate, Kondolf said. Some dams, such as Fall Creek Dam in Oregon, have mechanisms that enable engineers to release sediment. But in the past, reservoirs were not designed to manage sediment and little thought was given to how to decommission them.

"This is one of those issues that falls in the cracks between different agencies and responsibilities," Kondolf said. Some larger facilities fall under the remit of the federal government. Other dams fall under state governments. California, for example, recently launched a state-sponsored program to future-proof the state's dams.

But smaller dams often get overlooked, Kondolf said.

One option is to remove the dams entirely. For example, sediment-related concerns contributed to the decision to remove the Elwha and Glines Canyon dams in Washington. The Elwha River later carried most of the sediment to its river mouth.

For smaller dams with a relatively small amount of sediment, "you just blow up the dam and let the sediment work its way downstream", Kondolf said.

There are other technological solutions, but they are expensive, such as retrofitting dams to enable them to pass sediment; creating a bypass to route sediment-heavy water around the dam; and mechanical dredging.

Another option is to empty the reservoir of water and let the river cut a channel through the sediment, Kondolf said.

However, released dam sediment brings its own risks. If not managed carefully, sediment can flood the river system and disrupt downstream ecosystems.

"The economics of managing reservoir sediment long-term were not generally incorporated into dam design and planning in the last century when many western-U.S. water projects began," East said. This has "pushed the cost burden on future generations."

But the reality is that reservoirs around the country are filling with sediment. It is reducing their ability to store water and potentially endangering the communities living downstream of them — and unless addressed, the problem will only get worse.

When the famous Chicxulub asteroid struck what is now Mexico's Yucatán Peninsula 66 million years ago, it ended the reign of the nonavian dinosaurs. As ecosystems recovered, small, adaptable mammals rapidly spread into newly vacant niches, evolving new diets, larger bodies and more diverse ways of life until they dominated the land within a few million years.

Now, new evidence suggests that even before the asteroid hit, Earth was already beginning to rot.

A study published May 12 in the journal PNAS found evidence of environmental fungal blooms before and after the Cretaceous-Paleogene (K-Pg) mass extinction event. Looking at microscopic fossils from rocks dating to the K-Pg event in North America, a team of researchers led by Dr. Arturo Casadevall, chair of the Molecular Microbiology and Immunology Department at Johns Hopkins University, identified an earlier fungal bloom dating to roughly 30,000 to 100,000 years before the asteroid hit, during a period of cooling linked to intense volcanic activity in the Deccan Traps in what is now India.

The researchers also found a fungal spike immediately after the Chicxulub impact exactly when dead plants and animals would have littered devastated ecosystems. The finding adds to earlier evidence from New Zealand that suggests the fungal bloom wasn't just a local event.

This idea feeds into a controversial idea called the fungal infection-mammalian selection (FIMS) hypothesis. First proposed two decades ago by Casadevall and subsequently updated in 2018 and 2026 papers, the FIMS hypothesis suggests that the abundance of fungus before the asteroid impact may have helped mammals gain an evolutionary advantage over the dinosaurs that remained after that cataclysm.

The hypothesis doesn't argue that fungus wiped out the dinosaurs. Instead, it proposes that mammals' live births and immune systems may have helped them withstand the rise of fungus in ways egg-laying reptiles and nonavian dinosaurs could not, thereby giving them the upper hand in disrupted ecosystems full of rot, cold, starvation and polluted air.

"The fungal infection mammalian selection hypothesis is focused on how the mammals became the dominant land animals," Casadevall told Live Science via email, "not dinosaur extinction."

But not all experts are convinced. Mary O'Connell, the Chair of the Zoology department at the University of Manchester told Live Science via email that the hypothesis doesn't fit all the evidence and that "it emerges from a chain of plausibility arguments rather than direct evidence, and the links in the chain are weaker than the framing may suggest."

A fungal world before the asteroid

To find this pre-asteroid fungal bloom in their new study, Casadevall and Rosanna Baker, who researches fungus at Johns Hopkins, focused on palynomorphs — tiny fossilized organic remains, including fungal spores, fungal hyphae and plant pollen — preserved in ancient sediment layers from the Denver Basin in Colorado dating back to the K-Pg extinction.

Baker counted between 100 and 500 microfossils per sample across sediment layers spanning about 60,000 years before the K-Pg mass extinction event to 30,000 years after it. Most samples were dominated by plant material, but some layers contained 50% or more fungal spores. The researchers interpreted this high proportion of fungal spores as a sign of ecological disturbance ‪—‬ more dead material and more fungi feeding on it, especially after the dinosaur-killing asteroid.

Baker and Casadevall also found an earlier fungal bloom dating to roughly 10,000 to 30,000 years before the asteroid impact.

"What surprised us was the proliferation in association with Deccan volcanism, which implies ecological disruption prior to [the] meteor impact," Casadevall told Live Science via email.

In other words, ecosystems may have already been under stress — and faced with a fungal overgrowth — before the asteroid delivered the final blow.

The case for FIMS

When Casadevall first proposed the FIMS hypothesis, it rested on a simple observation: Mammals are unusually resistant to invasive fungal diseases.

Fungi are major pathogens of plants, insects, amphibians and many reptiles, but life-threatening systemic fungal infections are relatively rare in mammals with healthy immune systems. Casadevall argued that mammals' relative resistance to fungi came from their sophisticated immune systems. In mammals, first-line, relatively nonspecific defenses such as neutrophils recognize fungal cell walls, engulf spores and attack invasive filaments, while the adaptive arm of the immune system excels at building customized antibodies that selectively target specific fungi, and then creating immune memory so the invaders are recognized the next time they're encountered.

Casadevall also initially believed that mammals' high, stable body temperatures — generally around 97 to 104 degrees Fahrenheit (36 to 40 degrees Celsius) — could prevent many environmental fungi from growing well enough to establish an infection.

"The success of mammals made no sense given their high energy needs and this idea provided a plausible mechanism for their success," Casadevall said.

After the asteroid hit, forests burned, plants died, darkness fell and dead matter piled up, providing a feast for fungi. If fungal spores filled the air, surviving animals may have inhaled enormous, possibly dangerous doses.

However, the FIMS hypothesis proposes that mammals had several possible adaptations against these fungal invaders that other animals did not. In addition to having warm body temperatures and complex immune systems, mammals who could maintain stable body temperatures could forage during the cold "impact winter," ‪when ash and dust blocked out the light of the sun. Meanwhile the dinosaurs, who were originally thought to be the "cold-blooded," would have struggled to stay active and find food, leading them to become weaker and more susceptible to disease.

Differences in gestation may have mattered, too. Many mammals protect developing embryos inside the mother's body, where warmth and immunity could shield them from fungi. In contrast, all dinosaurs laid eggs. Those eggs developed in nests, which were often in contact with soil and decaying plant matter ‪—‬ the exact environments where fungi thrive. Modern fungi can invade reptile eggs, and fossilized hyphae, a type of fungus, was described in dinosaur eggshells in a 2008 study.

That does not prove fungi wiped out dinosaur nests. Fossilized hyphae could suggest either that fungi invaded a living egg, or that fungi colonized the egg after death. Still, eggshells provide one clue that fungi were a problem for the survival of dinosaur young..

A sick sauropod

Another potential line of evidence to buttress the FIMS hypothesis came from dinosaur bones that showed evidence of fungal infection.

In 2022, a team led by Cary Woodruff, curator of vertebrate paleontology at the Phillip and Patricia Frost Museum of Science in Miami, described a Late Jurassic sauropod fossil from Montana with unusual lesions in its neck vertebrae. The dinosaur, an immature diplodocine known as MOR 7029, lived roughly 150 million years ago, long before the asteroid impact.

Sauropods had long, air-filled neck bones connected to their respiratory system. In fossils of healthy sauropods, the places where air-sac tissue met bone are smooth, sometimes glass-like, Woodruff told Live Science.

But in this animal, the normally smooth sockets were rough and irregular. "Coming out of the socket — like imagine coming out towards you —it looked like fossilized heads of broccoli florets," Woodruff said. "And I had never seen that."

Unsure what he was looking at, Woodruff posted a photo online and asked whether anyone had seen anything similar. Other anatomists and paleontologists quickly weighed in. Their response, he recalled, was essentially, "That's exactly what we would have predicted a respiratory infection to look like," he said.

By comparing the fossil lesions with respiratory diseases in modern birds, which are living dinosaurs, the team tentatively flagged the condition as airsacculitis with associated osteomyelitis ‪—‬ air-sac inflammation that had spread to nearby bone.

However, the researchers could not identify the exact disease that caused the inflammation. In modern birds, airsacculitis can be caused by fungal infections such as aspergillosis, but also by bacterial infections. Without preserved soft tissue or fungal structures, they stopped short of confirming a fungal infection diagnosis.

Either way, Woodruff thinks the sauropod was probably sick enough that the infection contributed to its death. It may have become too weak to keep up, too ill to eat or drink properly, or too vulnerable to escape predators.

The case against FIMS

Even so, many experts remain skeptical of Casadevall's hypothesis. For one thing, the asteroid and its aftermath already explain a tremendous amount. The impact triggered tsunamis, earthquakes and a global food-web collapse. Those disasters alone would have killed off large dinosaurs at the top of disrupted food chains.

Another problem revolves around body temperature. In 2005, when Casadevall first proposed his hypothesis, dinosaurs were often thought of as "cold-blooded," while mammals were "warm-blooded." But over the years, research has uncovered that many nonavian dinosaurs, such as Velociraptor, T. rex and Brachiosaurus, were likely warm-blooded, too ‪—‬ so body temperature wouldn't explain why fungus would have been a big problem for many dinosaurs.

Complicating the picture, birds — the only surviving dinosaurs — often have higher body temperatures than mammals do. So the FIMS theory rests on an assumption that living examples of dinosaurs don't support, Flynn said.

Around the time of the asteroid impact, cold-blooded creatures like turtles and crocodilians that lived in the water "don't really suffer as severely as other groups, even including mammals," Andrew Flynn, an assistant professor of paleobotany at New Mexico State University, told Live Science. "Clearly this fungal thing is not probably what's causing mammals to rise up and go on to dominate the ecosystem."

The FIMS hypothesis also has a contradiction at its heart, said Jingmai O'Connor, a paleontologist at the Field Museum in Chicago.

"It's just a hypothesis I had vaguely heard about but was not popular, not often mentioned, and I guess I would say 'out of favor,'" O'Connor told Live Science via email. "If everything we know that suggests dinosaurs are more susceptible to fungal infections is based on living dinosaurs (birds) and they survived, doesn't that rather undermine this hypothesis?"

Dr. Isabel Jimenez, an assistant professor at Johns Hopkins University and a proponent of FIMS, agrees that this discrepancy must be explained.

"For the FIMS hypothesis to still hold water, we have to consider alternative mechanisms by which dinosaurs could have been more susceptible to fungal diseases than mammals," Jimenez told Live Science in an email.

Modern-day birds are avian theropods — the surviving branch of the dinosaur family tree — and likely faced the same fungal pressures as their extinct relatives. More recently, in 2026, Jimenez and Casadevall have suggested a tweak to their theory: that some of the small-bodied, beaked, flying or semiflying theropods may have been able to pull through, despite some baseline higher susceptibility to fungal disease. That's because they were small, had feathers to better insulate their bodies, and maintained generalist diets that enabled them to use more food sources than the dinosaur lineages that died out.

Working with Jimenez, Casadevall outlined an updated version of the FIMS hypothesis in a paper published earlier this year. In it, they propose that dinosaurs' respiratory systems may have made them more at risk than their mammalian counterparts.

Whereas mammals breathe by expanding and contracting flexible lungs filled with blood-rich alveoli, modern birds — and probably many dinosaurs — have rigid lungs ventilated by a network of thin-walled air sacs. The air sacs act as bellows, maintaining one-way airflow through the lungs, but they perform little gas exchange and have relatively poor blood supply. This system could have also carried inhaled fungal spores deep into the respiratory tract, where they may have settled in air sacs that are harder for circulating immune cells to reach and clear.

While birds have a similar sophisticated immune system as mammals — including macrophages and antibodies — their air sacs extend throughout the body and sometimes into the bones, making fungal infections easier to spread beyond just the respiratory system.

This possible weakness is what Jimenez and Casadevall suggest is what made air-sac-bearing dinosaurs vulnerable to fungal infections, similar to modern birds.

After the asteroid strike, there would have been soot and sulfur polluting the air and dinosaurs would have been nutritionally stressed. "So secondary fungal infections are certainly possible," Jimenez said, referring to infections that take root after the body's defenses are weakened.

But skeptics argue that even with these updates in the FIMS hypothesis, more evidence is needed to show it's not just possible but likely.

The idea that their air-sac respiratory systems made them more vulnerable to fungus is "a more testable claim," but its likely impact on dinosaurs would also be more limited, because of competing environmental factors, O'Connell said.

And because only a handful of fungal species actually cause disease in animals, to show that fungi worsened dinosaur deaths, researchers would need to find evidence of elevated fungal disease in dinosaur fossils from right after the asteroid struck, O'Connell said.

"We would also need to be able to identify elevation of pathogenic fungal spores in the sedimentary record at that boundary, rather than just more spores," she said.

Overall, "the fungal infection-mammalian selection hypothesis feels like a conjecture in search of any supporting evidence," paleontologist and science writer Riley Black, told Live Science in an email.

How much do you know about the king of the dinosaurs? Test your knowledge with our T. rex quiz!

LONDON — In the packed halls of the recent Alzheimer's Association International Conference in mid-July, over 10,000 experts in brain research discussed strategies to beat dementia. But in one room, attendees gathered to watch a video from a nonscientist named Isabel Beltre, whose life was transformed by a recent trial.

Beltre was one of 1,200 participants ages 60 to 77 who took part in the Latin American Initiative for Lifestyle Intervention to Prevent Cognitive Decline (LatAm-FINGERS), the first trial to test a culturally adapted lifestyle intervention to prevent cognitive decline in Latin America. Rather than attempting to treat dementia after it has developed, the trial tackled risk factors that raise the likelihood of the condition, like smoking and a sedentary lifestyle.

In her native Spanish, Beltre thanked both God and the project for its impact on her life. There was a celebratory air in the conference room as the trial's results were released. The lifestyle interventions, which included exercise programs and brain training, had statistically improved participants' thinking skills, compared with a control group who received only bare-bones support, like general health advice.

Trials like LatAm-FINGERS — including a related U.S.-based trial called the POINTER study — ultimately aim to cut rates of dementia around the world. The LatAm trial is the most successful of these trials to date, raising hopes that this approach could help to prevent dementia en masse.

Whether that promise pans out remains to be seen, though, experts told Live Science.

The lifestyle factors that increase dementia risk

When it comes to dementia prevention, few guidelines are as influential as those written by the Lancet Commission, an expert-led review of the available evidence on dementia. Three editions of this review have provided comprehensive overviews of risk factors; the most recent, published in 2024, identified 14 factors, including smoking, infrequent social contact, and a lack of physical activity.

The commission estimated that, together, the identified risk factors contribute to 45% of dementia cases at a population level.

Long before the 2024 report, Dr. Miia Kivipelto, a professor of clinical geriatrics at the Karolinska Institute in Sweden, wanted to tackle these risk factors at the individual and population level in her work as a medical doctor and epidemiologist.

"I felt that I wanted to move on to interventions to really show that if you do something, you can change the risk," Kivipelto told Live Science.

The many contributing factors that lead to cognitive decline, and ultimately dementia, couldn't be tackled in isolation, Kivipelto realized. She designed a trial in her native Finland, called the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) study, that would aim to tackle many risks simultaneously.

The two-year project recruited 1,260 participants ages 60 to 77 and divided them into two groups. The first group received general health advice, such as tips on healthy eating and exercise, at the start of the study and then again after six and 12 months.

The other group additionally received a multipart intervention consisting of meetings with nutritionists to discuss personalized meal plans, brain training, exercise programs, social activities, and metabolic and vascular tests. These programs lasted throughout the study period, with the brain training alone including 144 sessions over 12 months of the study.

At the end, the two groups' cognitive skills, such as thinking speed and memory, were tested; the trial did not track dementia rates, only overall cognition. In both groups, these skills improved by the end of the study. But the headline finding was that in the group given the comprehensive intervention, overall brain function was judged to be 25% stronger than in the control group.

What did that difference mean in concrete terms? It turns out, the improvement was relatively small.

It was calculated in terms of a statistical measure called a z-score, which represented the extent of the improvement in brain-function test scores in each group. In the end, the difference in z-scores between the two groups was just 0.04 points, which a 2021 review defined as "small."

However, Kivipelto maintained that the improvement is significant. "It's as big as anti-amyloid treatments," she said.

Several drugs approved by the U.S Food and Drug Administration for the treatment of Alzheimer's disease target amyloid, a protein that forms unruly tangles in the brain. These drugs have been reported to marginally slow the disease's progression, but statistically, their overall effect is small.

Targeting those most at risk of dementia

Despite criticism of the trial's moderate effects, Kivipelto's FINGER intervention has since been repeated all over the world, including by the U.S.-based POINTER study, and, most recently, LatAm-FINGERS.

Dr. Gill Livingston, a psychiatrist at University College London who led the 2024 Lancet Commission report, told Live Science that the data from these trials shows that dementia interventions need to be tailored to the risks the participants actually face.

In the POINTER study, 70% of the participants had a college degree, compared with roughly 43% of the wider U.S. population. "People who are usually in randomized controlled trials are usually highly educated people who are very interested in health," Livingston noted. "And that means they're at less risk of developing dementia, and therefore, less can be done to help them" through lifestyle interventions.

The difference between the test group and the control group in the POINTER study was even smaller than in the original FINGER trial — just 0.029 points.

LatAm-FINGERS tried to maximize flexibility in its approach by better tailoring its interventions to its study participants, said study first author Dr. Lucía Crivelli, a neuropsychologist at the neurology center FLENI in Argentina. Dietary advice reflected the produce available in each region. Exercise programs considered local environmental factors; for example, sessions were held outside, avoiding hot gyms during the sweltering summers in Ecuador and Mexico.

The trial recruited volunteers across Argentina, Brazil, Bolivia, Chile, Colombia, Costa Rica, Ecuador, Mexico, Peru, Puerto Rico, the Dominican Republic and Uruguay. The participants were less highly educated than the POINTER cohort had been; 37% had no education beyond a high school diploma, compared to just 5% in the POINTER study. That likely made the study participants more representative of people who might benefit from the trial's interventions.

When Crivelli announced the results of the trial at the conference, a large cheer filled the room. The intervention had achieved a difference of 0.11 points between the test and control groups' z-scores — more than twice the difference seen in the FINGER trial.

That difference corresponded to a relative 55% improvement in overall brain function in the test group compared to the control group at the end of the study.

What made the difference?

Crivelli, Kivipelto and Livingston all pointed to one key factor in explaining the project's relative success. "In Latin America, there are more modifiable risk factors," Kivipelto summarized.

The study participants had lower academic attainment and higher cardiometabolic risk factors, such as high BMI, than the participants in previous studies. The trial had reaped the rewards of delivering the intervention to people who needed it most, experts concluded.

But the approach wasn't perfect. said Dr. Claudia Suemoto, a neuroscientist at the University of São Paulo who presented a detailed breakdown of the trial's results. They found that the intervention didn't improve men's overall cognitive ability, compared with the control group. Only women benefited.

The difference may be that men simply didn't engage in the trial as much as women did, Suemoto's data suggested. Livingston pointed to the trials' exercise classes as an example of an activity that men may not be as likely to participate in, although that's an assumption rather than being based on the trial data.

Can we really prevent dementia?

The LatAm-FINGERS results show that addressing the dementia risk factors that people face is the most effective way to prevent dementia, Livingston said.

And Livingston is leaning into the idea. She's planning a dementia prevention trial in the U.K. that will ask participants about their risk factors and then give them tailored interventions based on the risks they can realistically avoid.

"We ask them to choose which ones they want," Livingston said. "If somebody doesn't want to stop drinking or smoking, then they're not going to."

While LatAm-FINGERS has been the most successful of the FINGER trials to date, it's still unclear whether such projects actually reduce dementia rates.

Only two previous prevention trials have attempted to measure dementia incidence, and neither saw a notable change in their test group. That said, neither trial included cognitive training interventions like FINGER did; they instead focused on cardiovascular risk factors, such as low exercise rates and high body weight.

Measuring changes in dementia rates requires trials much longer than two years, and they also need massive recruitment. Implementing an intensive study like LatAM-FINGERS in the wider community "is not practical," Crivelli said. Instead, she said the study's interventions would need to be distilled into their essential parts so that they could be implemented at scale.

Kivipelto is optimistic that advances in Alzheimer's screening, including new blood-based tests, could provide an early biological measure to show that FINGERS-style interventions can help stave off dementia. This would be an indirect measure but wouldn't require as long a trial as tracking dementia rates.

Some of the risk factors the Lancet Commission identified, like air pollution, can't be modified at an individual level, Livingston noted. A combination of personal and governmental change could address this, she said, as it has in the realm of infectious-disease prevention.

While the scientific field still has uncertainty about the scale of change that trials like FINGERS can bring about, Beltre has little doubt about the trial's impact.

While she previously relied on help from family members like her grandson to complete everyday activities, she said, "I do everything myself now."

This article is for informational purposes only and is not meant to offer medical advice.

See how much you know about the most complex organ in the human body with our brain quiz!

As recently as 2019, U.S. military health researchers had concerns about testosterone levels among the troops. Their worries were not that male service members didn't have enough of the hormone, as current Defense Secretary Pete Hegseth fears. Instead, experts worried about the troops' inappropriate use of testosterone replacement therapy (TRT), the practice of using injectables or gels to increase a person's circulating testosterone.

"Out of every 1,000 male service members, almost 3 are inappropriately receiving TRT," researchers wrote in a report published on Health.mil, the military's health system website. (That report is available on the Wayback Machine.)

This could lead to harmful side effects, the report authors noted.

"Those being inappropriately treated may experience adverse effects of the medication," including obstructive sleep apnea and edema, the report continued. TRT may worsen sleep apnea and cause swelling in the limbs from water retention. "These adverse effects have the potential to impact deployability and medical readiness."

In short, supplementing testosterone unnecessarily could reduce military readiness, rather than boost it, per the report authors. Now, in a reversal of these concerns, more service members might find themselves offered TRT in the wake of a new military policy announced July 15 by the Department of Defense. It states that all active-duty and reserve service members' testosterone levels will be tested annually starting at age 30.

It is unclear how the policy might affect female service members. Live Science contacted the Department of Defense, which declined to answer questions about the expected testing protocol or criteria for offering service members TRT.

Such an initiative goes well beyond what medical organizations recommend regarding hormone screening for men, experts told Live Science. And while the screening could catch some service members who actually need treatment, the policy might also lead to overtreatment and, in turn, unwanted side effects.

"One of my big, big concerns around this is, how are these tests going to be conducted?" said Luke Cox, a lecturer in sport and exercise science at Swansea University in the U.K. who studies performance-enhancing drugs, including steroidal hormones like testosterone.

Testosterone and military readiness

Testosterone has strong cultural connections with hypermasculinity. As the primary hormone responsible for the development of male sexual characteristics, it's been associated with concepts of strength, aggression and virility. Hegseth continued this pattern by tying the hormone to military readiness in a video announcing the policy change.

"[I]t's about restoring and optimizing your natural capabilities, projecting your longevity and ensuring you have the biological foundation required to sustain the fight," Hegseth said.

That rhetoric aligns with promises made in advertisements for TRT, which pitch testosterone as a tool to "attain that higher level of living," Cox told Live Science. The clinical trial data is more circumspect, however.

In men with genuinely low circulating testosterone levels — often defined as lower than 300 nanograms per deciliter (ng/dL) — TRT does have benefits. (This 300 ng/dL threshold isn't the universal definition for "low testosterone," as not everyone below that line has troubling symptoms. We'll unpack that later.)

Much of the clearest evidence of benefit is for sexual symptoms: In a large clinical trial, men with low testosterone saw increased sexual desire and sexual activity, such as masturbating or having sex, after TRT. For men whose levels are not low, additional supplementation hasn't been shown to have a clear benefit in the bedroom.

TRT has been shown to improve mood and reduce depressive symptoms in men with low testosterone, and in older men, it's also been shown to improve bone density. Boosting testosterone also boosts muscle mass in middle-aged and older men.

Most of this research has been done on men over age 40, because testosterone naturally declines with age. Low testosterone is rarer in younger men, although exact estimates of its prevalence vary. Estimates also vary among older age groups, but a large study called the European Male Ageing Study found that 0.1% of 40- to 49 year-olds had low testosterone, along with symptoms such as fatigue or low sexual desire. That number rose to 5.1% among those in their 70s.

In 2024, 42.8% of active-duty military members, including men and women, were 25 and younger. Only 8.3% were 41 or older. But the prevalence of low testosterone is generally on the rise, said Dr. Joshua Halpern, chief scientific officer and regional medical director at Posterity Health and a reproductive urologist at Northwestern University's Feinberg School of Medicine.

Studies suggest that testosterone levels are lower, on average, in young men today than they were in the year 2000. Scientists have theorized that this trend may stem from increases in body mass index, as well as other factors, such as sedentary lifestyles, changing diets and exposure to environmental toxins, according to a 2020 article in Urology Times.

"While active service members are typically among the healthiest of men, and therefore at lower risk for testosterone deficiency, a significant proportion of service members could have a low testosterone number," Halpern told Live Science. "But that doesn't necessarily mean they should receive treatment."

Treating with testosterone

There is no universal testosterone level that is definitively "too low." The cutoff of 300 ng/dL is based on the point at which symptoms of low testosterone ‪—‬ such as fatigue, low libido and poor mood ‪—‬ tend to present themselves. However, some men have no problems associated with low or low-to-normal numbers.

Plus, testosterone doesn't act in isolation. In fact, some symptoms of "low T" in aging men may actually arise from low estrogen, which stems from these individuals' declining testosterone levels. That's because the male body converts a portion of its total testosterone into estrogen. The two hormones normally work together to regulate sexual desire and erectile function, while estrogen specifically regulates fat accumulation, studies suggest.

There is also a broad range of "normal," running from around 300 to 1,000 ng/dL. For that reason, the American Urological Association recommends testing only if a man is experiencing symptoms that point to low testosterone.

"Making a diagnosis of testosterone deficiency in the absence of signs and/or symptoms increases the likelihood of making a false diagnosis and reduces the potential benefit of testosterone therapy," the association's guidelines state.

Testosterone also has a complicated role in the body. It can affect mood, energy and bone health, but the hormone itself can also be affected by health conditions and lifestyle factors. Studies suggest that intense periods of physical training can temporarily suppress testosterone in military members. That finding tracks with civilian studies that have found that levels can shift dramatically based on people's recent sleep and exercise.

That's why doctors recommend two separate blood tests, both taken in the morning, before diagnosing low testosterone. This helps to ensure that a person's levels are consistently in a low range.

In some cases, it may be that the best treatment isn't TRT but rather an approach that addresses the underlying health conditions that are affecting testosterone levels. This might include assessing whether a patient has sleep apnea and treating it if they do, for example.

Meanwhile, pushing testosterone levels too high can lead to erythrocytosis, an overconcentration of red blood cells that can bring a risk of blood clots and stroke. What constitutes "too much" varies person to person. Due to some testosterone being converted into estrogen, too-high levels can also lead to unwanted side effects like breast growth in some men on TRT.

The other major side effect of TRT is that testosterone replacement shuts off the body's own production of testosterone in the testes. TRT only raises the level of testosterone circulating in the blood, not in the testes where sperm production takes place. Sensing the testosterone in the bloodstream, the brain suppresses its production in the testes, thus diminishing or stopping sperm production. This can cause infertility, and sperm production may take several years to recover after a person stops TRT.

"[F]or young service members who are found to have low testosterone, it will be very important for them and their physicians to understand the risks of treatment and the alternatives that may be more appropriate for preserving their fertility," Halpern said.

Meanwhile, a federal judge has questioned the treatment program. An ongoing lawsuit against the federal government's ban on transgender troops states that "the Armed Forces must adhere to high mental and physical health standards ... without the benefit of routine medical treatment or special provisions." The judge raised questions about how the TRT that will be offered to cis men differs from that which transgender men would use.

This article is for informational purposes only and is not meant to offer medical advice.

When OpenAI recently revealed that two of its most advanced artificial intelligence (AI) models had escaped the confines of a cybersecurity test and hacked into a startup, it sounded a lot like the kind of scenario that AI safety researchers have spent years warning about.

The models found a previously unknown vulnerability in the infrastructure meant to contain them, gained access to the public internet and broke into Hugging Face, a major platform for hosting AI models and datasets. Their objective, however, was less sinister than the sequence of events might suggest: They were looking for information that would help them complete the cybersecurity test OpenAI had given them.

In a July 16 statement, Hugging Face representatives disclosed that internal datasets had been infiltrated, saying it was "different from anything we had handled before" because it was driven "by an autonomous AI agent system." In another statement published July 21, OpenAI representatives fessed up to being responsible, calling the episode an "unprecedented cyber incident" while warning that similar events could become more common as AI models become increasingly capable.

Independent experts who spoke with Live Science agree that what happened is significant — but they cautioned against interpreting it as an AI system suddenly developing a malicious agenda. The models appear to have pursued the task OpenAI gave them, finding a route to success that their creators had failed to anticipate or adequately block.

"If there's a failure here, it isn't that the AI wanted to hack something," Oli Buckley, a professor in cybersecurity at Loughborough University in the U.K., told Live Science. "It's that humans created a test where success was measured by achieving an objective, deliberately relaxed some of the normal security controls to measure the system's capabilities, and underestimated how effective the model would be at finding an unexpected path to success."

How did an OpenAI test end up hitting Hugging Face like this?

OpenAI was testing GPT-5.6 Sol and a more powerful unreleased model using ExploitGym, a benchmark that challenges AI systems to find and exploit software vulnerabilities. The company removed some cybersecurity safeguards that would normally prevent potentially dangerous actions while relying on an isolated environment to keep the models away from the wider internet.

According to OpenAI's postmortem, the models discovered a previously unknown vulnerability in third-party software used to proxy and cache software packages. They exploited it, escalated their privileges and moved through OpenAI's research infrastructure until they reached a machine with public internet access.

Hugging Face became a target because the models identified it as a possible source of information that could help them complete the ExploitGym challenges. OpenAI said at least one attack chain involved stolen credentials and previously unknown vulnerabilities that eventually enabled the models to execute remote code on Hugging Face systems and access test solutions stored in a production database.

In their disclosure, Hugging Face representatives said the company recorded more than 17,000 actions during the intrusion, but they couldn't initially explain who or what was behind it. OpenAI's subsequent disclosure supplied that missing piece: Its models had broken out of their test environment and gone looking for the answers elsewhere.

Did the AI really "escape"?

It's notable that the models found a flaw in the infrastructure designed to contain an AI and used it to reach the public internet. Describing the models as having "gone rogue," however, risks assigning them unsupported motivations, Buckley said.

"I think I'd be wary of jumping to "rogue AI,"" Buckley said. "The models didn't develop their own agenda or decide to attack Hugging Face while twirling their digital moustache."

Buckley compared it to asking a dog to fetch a ball while leaving the garden gate open. "If the easiest ball for it to find is in the park down the road, that's where it'll head," he said. "You wouldn't say the dog had gone rogue; you'd just say you underestimated how literally it would pursue the task."

Daniel Hulme, entrepreneur in residence at University College London and CEO of AI safety company Conscium, agreed that the models shouldn't be assigned human-like motivations. "Models don't have intent; humans have the intent, and we train models with goals in mind," he told Live Science

The capability may matter more than the motive

What matters more than the models' supposed motives is what they managed to accomplish while pursuing their assigned task.

"The genuinely significant point is that the models appear to have chained together multiple vulnerabilities across different systems and sustained a complex sequence of actions," Buckley said. "That demonstrates a level of capability that security professionals should take seriously."

Katerina Mitrokotsa, a professor of cybersecurity and applied cryptography at the University of St. Gallen in Switzerland, said the containment failure is particularly concerning because another company ultimately paid the price.

"What concerns me most is who ended up affected," Mitrokotsa told Live Science. "The victim was not the company running the test, but a third party. This is the scenario security researchers have warned about for some time: that an AI agent's escape does not necessarily stay contained to the environment in which it originated."

OpenAI representatives said they have tightened the infrastructure used for these evaluations. But Mitrokotsa warned that containment becomes harder to guarantee as models improve at performing exactly the kind of exploitation OpenAI was testing.

An AI warning — and an impressive product demonstration

There is also reason to look carefully at how the incident is being framed. OpenAI's account serves two purposes at once: It warns about the security risks posed by increasingly capable AI while demonstrating just how capable its own newest models have become.

Buckley said announcements from frontier AI companies like OpenAI or Anthropic should be viewed in the context of an industry competing to build ever-more-powerful models.

"We've seen similar high-profile capability demonstrations from Anthropic and others," he said. "That doesn't make the findings untrue, but it does mean we should separate the technical evidence from the marketing narrative."

These companies have every incentive to show both that their models are extraordinarily capable and that they are taking the risks seriously, he added. The Hugging Face incident demonstrates both that OpenAI's models carried out a complex series of operations with considerable autonomy and that its security measures failed to keep them inside the experiment.

Hulme argued that the longer-term challenge is ensuring that increasingly capable AI systems pursue their goals in ways that remain consistent with human values.

"Rather than seeking to control AIs, the focus should instead be on alignment," he said, adding that continuous testing will be needed to ensure systems remain aligned with their intended missions while staying secure.

The episode, the experts said, leaves OpenAI with a result that is impressive and uncomfortable in equal measure. Its models found previously unknown vulnerabilities and continued pursuing their goal well beyond the boundaries their creators expected, but none of that requires them to have developed malign intentions.

"The lesson isn't that AI has become malicious," Buckley said. "Instead, it's that increasingly capable systems will exploit opportunities that humans fail to anticipate."

In this incident, OpenAI's new models were given a hacking challenge and they were rewarded for finding a way to solve it. The humans running the experiment simply hadn't anticipated quite how far they might go.

One of the central assumptions of modern neuroscience research is that the brain's shape and structure affect behavior. Scientists have linked a thicker cortex (the brain's outer layer) to higher intelligence, certain brain wave patterns to better volleyball ability, and higher connectivity between parts of the brain to chess-playing skills. There's a vast number of these so-called brain-wide association studies (BWAS).

But when experts repeat these studies, they can't replicate the results.

This "replication crisis" suggests that a substantial amount of brain-behavior imaging research rests on a shaky foundation. There are myriad problems affecting this area of research, said Randy Ellis, a senior scientist at Oracle who wrote about these issues while working as a biomedical informatician at the Icahn School of Medicine at Mount Sinai in New York.

Some of these issues aren't unique to neuroscience. They begin with what Ellis called the "original sin" of science: Academics are put under huge pressure to publish positive research findings.

But some of these factors do apply specifically to this field.

The problems are big enough that they are "halting the growth and development of science and the curing of diseases," Ellis told Live Science.

Tiny differences, small samples

Several brain studies that could not be reproduced in follow-up work show how problems can creep in. Each of the original papers has over 500 citations, with the number of citations reflecting how much these studies may influence thinking in the field.

For example, a landmark study in 2007 found that the brains of kids with attention-deficit/hyperactivity disorder took longer to mature. But a study published earlier this year found that this link vanished once the different rates of aging between boys and girls were taken into account. "That's what made the whole house of cards topple," Matthew Albaugh, co-author of the replication paper and a clinical neuroscientist at the University of Vermont, previously told Live Science.

In 2011, researchers at University College London published a study that found that the density of gray matter, the part of the brain where brain cell bodies are found, in several brain areas was linked to participants' number of Facebook friends. (In 2011, this was a vital social metric.)

A 2015 paper failed to replicate this finding, but the authors of the original paper argued that was because the follow-up study didn't follow the same protocol.

"It's hard to do exactly what was done in the original studies," said Sarah Genon, a neuroscientist at the Research Center Jülich in Germany who was not involved in either the Facebook study or its replication.

So, in 2019, Genon and her colleagues tried a different type of replication analysis.

They used a vast database and conducted more than 10,000 analyses of MRI scans from hundreds of volunteers. The median neuroimaging study sample size was around 25, so they divided this larger dataset into smaller ones and then looked for significant associations within each.