Last year, when astronomers announced the atmosphere of a distant planet, K2-18b, contained promising signs of life, the claim generated media headlines calling it the “ strongest evidence so far ” of life beyond the Solar System. But many astronomers did not find the evidence persuasive .

Now, a paper published today in Nature Astronomy quantifies that skepticism. In a survey of 496 astrobiologists sent out eight days after the original claim, only 6.6% agreed or strongly agreed that “scientists have probably found extraterrestrial life on exoplanet K2-18b.” The authors say the study shows how quick but wide-reaching surveys could be deployed to take the pulse of a scientific community, helping inform public understanding of scientific questions.

The effort is “supervaluable and important,” says University of Cambridge psychologist Sander van der Linden, who studies disinformation. Behavioral science shows people make judgments based on their perception of scientific consensus, he says, and so the rapid polls could help protect the public from potential misinformation.

The poll is one of the first few outputs from the Centre for Scientific Community Opinion Polling and Evaluation (C-SCOPE), which has spent the past 3 years refining its methods to rapidly measure scientific consensus, says founder Peter Vickers, a philosopher of science at Durham University, where the institute is based. C-SCOPE has developed a network of representatives at 80 institutions around the world, who help design and administer the surveys.

Astrobiology has provided a useful playground for C-SCOPE to test its survey procedures. Last year, the team found that in a sample of 521 astrobiologists, 86.6% agreed or strongly agreed that “it is likely that extraterrestrial life (of at least a basic kind) exists somewhere in the Universe.” The team also sent out a similar survey about evidence for ancient extraterrestrial life on Mars, finding that, compared with the K2-18B results, more astrobiologists—15.1%—agreed the evidence was persuasive.

The results are “remarkably in line with my own personal scientific opinion after having read the papers and discussed them with colleagues,” says Sara Seager, an astrobiologist at the Massachusetts Institute of Technology. Quantifying a community’s opinion is helpful, she says, “compared to me just saying, ‘Hey, no one believes this.’”

C-SCOPE’s first proof-of-concept survey cast a broader net. In June 2023, the institute contacted more than 20,000 scientists in 12 countries, asking them to respond to a single statement: “Science has put it beyond reasonable doubt that COVID-19 is caused by a virus.” They received 6807 responses—a much higher response rate than typical for surveys—with 93.2% of respondents agreeing or strongly agreeing with the statement, the authors reported in a 2024 PLOS One paper. The figure may seem low for such an uncontroversial statement, Vickers says. But the survey sample was very broad, including chemists, physicists, and earth scientists as well as biologists and health scientists. Many scientists chose the “neutral” option, suggesting they did not feel they had the relevant expertise.

This shows how sending a survey to too many kinds of scientists could undermine its results, says John Cook, a psychology researcher at the University of Melbourne. Consensus in climate science, for instance, is higher when focusing on climate scientists specifically rather than all earth scientists. Still, Vickers thinks it’s generally worth surveying widely, casting a net across disciplines or broadly within a discipline. Narrowing the pool of experts risks biasing the sample, which could undermine trust in the results, he says. It also risks missing minority opinions that may eventually become mainstream.

Rapid polling is a “great idea,” Cook says, but slower, more laborious methods to ascertain consensus are still beneficial. For example, Delphi processes, which generate multiple rounds of feedback on a set of statements to gather and refine expert opinion, have the benefit of actually bringing people to a point of agreement, says Temple University philosopher of science Miriam Solomon. “When you put people together and make them talk to each other, they change their views sometimes.”

At some point, C-SCOPE will branch out into more politically contentious topics, including climate change, Vickers says. The group may, for example, track changes over time in scientists’ beliefs about the likely level of global warming by 2100. He also sees potential in rapidly polling scientists on topics in the headlines—such as the claim that Tylenol taken during pregnancy causes autism —possibly generating results in as little as 2 days. But his team would have to take great care over the question they ask, he says, to preserve their neutrality and trust in the results. “We want to remain squeaky clean.”

Citing concerns that artificial intelligence will make it easier for anyone to build biological weapons, the leaders of several major AI companies—in a rare moment of unity—have penned a new letter urging U.S. lawmakers to impose tighter controls on firms that sell synthetic, made-to-order strands of DNA.

“AI systems are improving rapidly, and alongside incredible benefits to science and medicine, there is a real possibility that the knowledge barriers which have historically prevented bad actors from obtaining biological weapons will meaningfully erode,” states the 3 June letter , which is signed by the heads of OpenAI, Anthropic, Google DeepMind, and more than 50 other prominent players in AI, biotechnology, and national security.

The letter calls on Congress to pass legislation that would require companies that sell synthesized DNA and the machines that make it to carefully vet orders and customers, and to keep detailed records “so that any threat that might evade initial screening can be traced back to its source. … Awareness of traceability itself deters misuse.”

The push for regulation comes amid growing concerns that AI products, including large language models and specialized tools trained on troves of biological data, could enable nonspecialists to gather sophisticated information on how to construct deadly toxins or assemble deadly bacteria, viruses, or other pathogens , using equipment and techniques that are becoming cheaper and easier to acquire . Together, the combination could make for potentially catastrophic risks, such as an AI-designed pathogen that sparks a global pandemic.

It’s not a new concern: For decades, some researchers and security experts have called for tighter controls on synthetic biology and related fields. Those calls have resulted in voluntary agreements by many companies that synthesize DNA to screen orders for potentially problematic sequences, and by some research funders to require grantees to request DNA sequences only from companies that do vetting.

Now, such screening should become mandatory for all U.S. firms, say the signers of the letter, which also include the heads of many of the major DNA synthesis companies. And Congress should act soon, they add: “Given the pace at which the underlying technology is changing, we believe the need is urgent.”

In the U.S. Senate, a bipartisan group of lawmakers has already introduced legislation, the Biosecurity Modernization and Innovation Act , that would give the Department of Commerce 1 year to develop new screening rules. The letter issued today could help give that legislation a push, says Hayley Anne Severance, deputy vice president of NTI | bio, a nonprofit that works to reduce the threat of biological weapons. “It’s really important for us to bring this over the finish line … to prevent the misuse of DNA synthesis,” she says. (One of the letter’s signers is Christine Wormuth, president and CEO of the Nuclear Threat Initiative, NTI | bio’s parent group.)

But regulating U.S. DNA companies should be just one part of a larger effort to blunt the bioweapon threat, Severance says. To be truly effective, she adds, other nations with significant DNA synthesis capabilities, such as China, would also have to adopt similar mandatory screening rules .

The AI industry should also take steps to ensure its products aren’t enabling the threat, Severance says. For instance, firms could program chatbots to refuse queries for “research and laboratory protocols that would allow nonexperts to more effectively manipulate biological agents,” she says. And they could limit access to specialized tools used to design proteins and chemicals. But the challenge, Severance notes, will be “making sure good science is able to proceed while efforts to cause harm with biology are identified and stopped. It’s a balancing act.”

Some AI companies say they are already taking steps to prevent the misuse of their products by bad actors. But the letter, which was first reported by The Wall Street Journal , is silent on whether any new biosecurity regulations should also extend to AI firms.

Continuing its pushback against President Donald Trump’s proposed deep cuts to federal science spending, a House of Representatives spending panel today released a draft 2027 spending bill that would give the National Institutes of Health (NIH) $47.3 billion, a slight $100 million boost.

Although smaller than the 40% cut he wanted last year, Trump’s budget proposed cutting NIH by 12% in the fiscal year that begins on 1 October and eliminating some institutes. The bill from the House Committee on Appropriations subcommittee that oversees health agencies rejects both those proposals. But the measure contains other provisions of concern to the biomedical research community, including limits on overhead payments, animal studies, and pathogen research.

"The House’s proposed increase for NIH is a welcome step in the right direction and an important recognition that investing in medical research means investing in better treatments, new cures, and healthier lives for Americans,” says Erika Sward, executive director of United for Cures, a coalition of patient advocacy groups.

The $47.3 billion for NIH’s base budget includes slight increases for most of the agency’s 27 institutes and centers. It holds steady the $1.5 billion Advanced Research Projects Agency for Health, an independent agency within NIH, which Trump wanted to cut by 37%. The measure also ignores Trump’s proposal to eliminate three NIH components—the minority health institute, complementary medicine center, and Fogarty International Center—and merge the drug abuse and alcoholism institutes.

However, the budget for the Centers for Disease Control and Prevention would fall by $1 billion, or 11%, to $8.1 billion. Trump had proposed a 41% cut to the public health agency that has been battered by staff cuts and other upheaval in the past year.

The bill would also ban NIH funding for “gain-of-function (GOF) research.” That term is often used to describe studies that give a pathogen new capabilities that could make it more harmful to people. Some conservatives and scientists think GOF experiments on viruses created SARS-CoV-2, and the Trump administration is finalizing new restrictions on such research.

But the bill language describing the ban does not limit it to certain experiments and could sweep up not only low-risk pathogen studies, but “whole swaths of biomedical research,” says Allen Segal, chief strategy and public affairs officer for the American Society for Microbiology. “It seems more designed to provide a message than to make a serious attempt at addressing biosafety and biosecurity.”

The bill also includes language that would bar funding for painful biomedical research on cats and dogs. It’s unclear how many projects this would affect or what research could continue, given that the legislation allows exceptions for studies where no alternatives are available or where the animals could be put up for adoption when studies end. Still, “This is the strongest language I’ve seen in my years tracking this stuff that’s made it this far in the [legislative] process,” says Naomi Charalambakis, director of communications and science policy at Americans for Medical Progress, a biomedical research advocacy group.

“We’re thrilled … because NIH funds more testing on dogs and cats than any other federal agency,” says Justin Goodman, senior vice president of the White Coat Waste Project, an animal advocacy group that worked with lawmakers to introduce the text. The group is also pushing for similar restrictions in funding bills for other agencies.

Finally, the House measure would limit to 30% how much of an NIH grant could be used to pay for the overhead costs of research at a few dozen private institutions with large endowments. It’s hard to tell whether the limit would result in a net loss in these so-called indirect costs for these universities because the bill appears to use a different formula than the government currently does to calculate rates, and individual rates negotiated with an institution vary. At the same time, 30% is twice the 15% cap that several federal agencies planned to adopt last year before federal courts ruled they had acted illegally and Congress passed 2026 spending bills that preserved the current system.

The House subcommittee takes up the bill tomorrow morning, and the full Appropriations Committee on 9 June when an accompanying report is expected to be released. The Senate Committee on Appropriations has not yet introduced its version of the bill, and both chambers will need to work out a compromise bill in the coming months.

Sward’s group hopes any final version will include language limiting NIH’s use of multiyear grants, which slashed the number of new grants funded last year and will likely do so again this year.

The NIH spending bill continues a trend of Congress rejecting Trump’s proposed deep cuts to federal science agencies. In April, for example, House appropriators gave the National Science Foundation, NASA, and several other research agencies more money than the White House requested for 2027, although the agencies would still see cuts .

U.S. health agencies are moving to launch new research—and potentially pass regulations—on a food category that industry and some scientists have long dismissed as misleading: ultraprocessed foods (UPFs).

On 14 May, the National Institutes of Health’s (NIH’s) Council of Councils endorsed a $150 million initiative, yet to be publicly launched, to investigate how diets dominated by UPFs harm children’s health and how to help children avoid those foods. The Food and Drug Administration (FDA), meanwhile, is drafting a legal definition of UPFs as a step toward regulating them. “Once we get that definition of ultraprocessed foods, then we go to mandatory front-of-package labeling” to discourage people from consuming them, Health and Human Services Secretary Robert F. Kennedy Jr. told a congressional committee on 16 April. The recent moves signal the Trump Administration’s embrace of a term that has become both a focal point of the Make America Healthy Again (MAHA) movement and a source of scientific controversy.

“I really don’t like the term,” says nutrition scientist Susan Mayne, former director of FDA’s Center for Food Safety and Applied Nutrition who is now at the Yale School of Public Health. The definition largely ignores the nutritional makeup of food, she says. “We have decades of research on the nutritional attributes of products” and their effects on health. “If we ignore those decades of research, that’s not going to be helpful.”

But others think the term has its uses. Kevin Hall, a former NIH scientist who carried out widely cited studies of UPF-rich diets, says the UPF category offers “complementary lens” on food that recognizes factors such as convenience and corporate marketing. The UPF idea “is a compelling narrative that really captures people’s attention in a way that focusing on salt, sugar, and fat never had,” he says.

Epidemiologist Carlos Augusto Monteiro at the University of São Paulo coined the term in 2009 as part of a Brazilian food classification system called NOVA. The UPF label is not based on any direct measure of food processing, which is hard to quantify. Instead, Monteiro and his colleagues labeled foods “ultraprocessed” if they contained manufactured ingredients such as emulsifiers, flavors, protein isolates, and modified starches that aren’t found in home or restaurant kitchens. Those chemicals indicate a product was born in the food industry’s laboratories, with the power of corporate marketing behind it. Monteiro says such ingredients are “essential to the success of the UPF business model.” UPFs are estimated to make up more than half of the calories people in the United States consume.

Food activists quickly adopted the term, but many nutrition specialists have been skeptical. Foods that most nutritionists consider beneficial, such as whole-grain bread, are classified as ultraprocessed if they contain dough conditioners or preservatives.

Yet researchers have linked UPF-dominated diets to indicators of adverse health effects, such as obesity, heart disease, and diabetes. A study released this week in the American Journal of Public Health found an association between higher consumption of UPFs and increased rates of dementia and cognitive impairment in older adults.

They also seem to promote overeating. In one of Hall’s experiments, study participants received either UPF-rich or minimally processed meals. Both alternatives contained matching amounts of calories, sugar, sodium, fat, and fiber. When offered the UPF-rich diet, people ate more food, consumed more calories, and gained more weight . Specific aspects of the UPF-rich diet seemed to be responsible: the higher calorie density of many UPFs, or their ability to activate reward systems in the brain that make a food “hyperpalatable.”

Critics of the UPF concept say it’s these narrower characteristics that deserve attention, rather than the broad “ultraprocessed” category. In a commentary published today in Science , Faidon Magkos at the University of Copenhagen and his co-authors argue food policies should focus on whether foods are “nutritionally poor, calorie-dense, and rapidly consumable,” rather than whether they fit the NOVA definition of ultraprocessed.

NIH’s proposed initiative would fund five randomized, controlled trials to learn exactly how particular components of UPFs might contribute to chronic disease. Other studies will examine how well particular interventions, such as outreach in schools, work to cut UPF consumption among children.

One driver of the recent momentum seems to be politics. A poll published by Politico in April found one-third of all people in the U.S., and half of all 2024 voters for President Donald Trump, described themselves as supporting the MAHA movement—and nearly 90% of MAHA supporters were in favor of removing UPFs from U.S. diets. More than half said they consider this a “core principle” of MAHA. No other policy goal had more support.

Hall says this groundswell of MAHA hostility probably helped inspire NIH’s new initiative—which he welcomes. Scientific research “is a social enterprise, and part of the way you get things done is to get people on board,” says Hall, who resigned from NIH in 2025 over what he called censorship by NIH leaders.

Yet some doubt the Trump administration’s public opposition to UPFs will lead to actual restrictions. “The administration talks big when it comes to taking on Big Food, but the actions often don’t match the rhetoric,” says Eva Greenthal, a senior policy scientist at the Center for Science in the Public Interest (CSPI). Mayne says the administration is caught between MAHA activists’ demands and food companies’ desire to avoid an off-putting label, which may explain why Kennedy’s promised definition of UPFs has yet to emerge.

Meanwhile, several outside groups have stepped forward with their own definitions. Healthy Eating Research, based at Duke University’s Global Health Institute, assembled an expert panel that released its proposal in May and presented it to officials at FDA and the White House. It follows NOVA’s original, expansive definition of UPFs, but would exempt most processed foods that qualify as “healthy” under criteria that FDA released near the end of former President Joe Biden’s administration.

CSPI is proposing a narrower definition that targets foods containing a shorter list of additives—those for which there’s evidence of harm, such as the preservative sodium nitrate. “We want to put something forward that is most likely to withstand the pretty much inevitable legal challenge from industry,” Greenthal says. CSPI also wants to include foods loaded with calories or salt even if they’re made from old-fashioned kitchen ingredients—meaning its definition would cover many brands of potato chips.

Former FDA Commissioner Robert Califf, who led the agency for parts of the administrations of Biden and former President Barack Obama, says any regulation targeting UPFs faces big hurdles. “It’s really hard for the FDA, at least in normal times, to make a regulation when there is substantial disagreement among experts.”

Picture reaching for something on a high shelf and coming up short. You don’t give up; instead, you find a stool, carry it over, and climb up. Buried in that action is something remarkable: You held the goal in your mind, identified what you needed, and executed a plan. No training required. A study published today in Science suggests bumble bees can do the same— the first demonstration of this kind of goal-directed problem-solving in an insect .

The finding “adds another level of complexity to what bees can do,” says Gema Martin-Ordas, a comparative psychologist at the University of Stirling who was not involved in the work. “It’s a really interesting line of research that will open up new questions.”

Previous studies have shown insects can learn to use objects to obtain rewards. In string-pulling experiments, for example, bees can learn to drag food within reach by tugging on a tether. But the behavior is trained, and the insects could have landed on the solution by watching others.

For the new study, researchers aimed to see whether bees could generate a solution independently by connecting two separate pieces of knowledge without ever being shown how to combine them. To do this, they set up a circular arena with a blue artificial flower mounted on the ceiling and a small Styrofoam ball placed nearby. To reach the flower and collect a sucrose reward, a bee had to roll the ball across the arena floor to the flower’s location and use the ball as a steppingstone to reach it. The insects were never shown that the ball could serve that purpose—the question was whether they could figure it out on their own.

To establish what prior experience bees actually needed, the opening experiment divided bees into three groups: those familiar with both the ball and the flower, those familiar with only one, and those with no prior exposure to either. Only the bees with exposure to both the flower and ball solved the task—rolling the ball to the flower to claim the reward.

In humans, this is like being handed a hammer and nail for the first time: If you’ve never seen either, you’re lost. If you’ve only seen one, you’re still stuck. But knowing what both do separately, there’s a chance you can work out how they fit together.

The team then needed to confirm the bees were acting deliberately, not just getting lucky. Because bumble bees are known to roll balls on their own even without any reward, a bee could theoretically push the ball to the right spot entirely by chance.

To rule that out, the final experiment forced a decision. Bees entered a rectangular arena containing two visually occluding compartments. The team hid a flower behind an opaque wall on one side. Each bee was first allowed to observe which side concealed the flower; then, once testing began, that view was blocked. The flower remained in place, and its sucrose reward with it, but the bee could no longer see it from where the ball sat. To claim the reward, the bee had to remember the correct side, return to the ball, and roll it there without any visual guidance, much like a person who spots something across a room, leaves to find a tool, and returns to the right spot from memory alone.

Most of the bees—23 out of 30—chose correctly on the first try. The only behavioral variable that predicted success was how long the insect spent inspecting the flower-side compartment before going for the ball. That pause, the researchers suggest, may have been something closer to the pause humans take when mentally rehearsing a plan before acting on it. “One possibility is that there’s some kind of mental restructuring happening, where they’re piecing together pieces of information,” says lead author Akshaye Anand Bhambore, a behavioral ecologist at the University of Oulu.

Lars Chittka, a behavioral ecologist at Queen Mary University of London who was not involved in the study, calls the design particularly convincing because the researchers had full control over what the bees had and had not experienced before testing. In earlier work on insect cognition, he says, it was difficult to rule out that animals had absorbed a shortcut during training. “Here, that was clearly under control.”

The findings add to a growing body of evidence that sophisticated cognition is not limited to animals with large brains—with implications for how scientists and conservationists think about insect welfare. For Chittka, the stakes are straightforward. “A new motivation for bee conservation,” he says, “arises out of respect for their mental abilities—from the observation that they are, at some level, thinking and possibly feeling beings.”

One afternoon in April, Cecilia Garraffo settled down at the head of a conference room table in Cambridge, Massachusetts, and gazed out at what might be the last astrophysicists of their kind.

The walls of this room had, in the past, reverberated with the din of thousands of other groups of scientists. Now, as streaks of sunlight poured in, the discussions turned to nonhuman collaborators. One by one, the gathered researchers discussed how they planned to apply machine learning to problems in astronomy. Observing an interstellar comet. Discerning wispy filaments of galaxies at the universe’s largest scales. Developing a new “tokenizer” that can translate astrophysical images into a form more readable by artificial intelligence (AI). “Sometimes models will be overconfident,” Garraffo warned a junior team member.

Afterward, as everyone filed out, black hole researcher Daniel Palumbo made a brief announcement. Representatives from AI chipmaker NVIDIA were on campus in search of scientists who wanted to solve problems using their hardware. To anyone who might need extra processing power, “today’s the day,” he said.

The Center for Astrophysics | Harvard & Smithsonian employs more than 600 astronomers and other staff, making it one of the world’s largest concentrations of professional stargazers. Garraffo heads its AstroAI group, which is charged with leading the center’s approach to applying machine learning to various problems. In just 4 years since they first proposed this specialized team, Garraffo and her colleagues have forged collaborations throughout the building and with industry teams such as Google DeepMind and Anthropic.

Originally, their goal was to use machine learning and AI to remove the technical barriers of math and computation while preserving what Garraffo considers the fun part of physics: honing scientific questions. Their toolbox did not include chatbots. Despite the buzz around ChatGPT, which was released within months of AstroAI’s first proposal, Garraffo had thought her group would steer clear of the headline-grabbing tool and other large language models (LLMs).

At least, until recently.

Now, stories of miraculous progress were starting to spread across the institution. As her MacBook Air pinwheeled to a crawl thanks to an AI agent running software locally, Garraffo’s colleague Alyssa Goodman showed me a data-fitting problem. She wanted to understand how the spiral arms of a distant galaxy were moving. But isolating just that motion from other patterns imparted into her data by the spin and the geometry of our own Galaxy had thwarted her group for years. She asked ChatGPT, which resolved the problem in a few minutes. Now, her research group was planning to write several papers on the resulting data set, “the single best map of spiral arm kinematics ever—like, by a factor of 100.”

Conversations comparing these tools with human researchers, once an underground whisper, have grown into a deep rumble in astrophysics departments around the world. Many are turning over aspects of their research practice—searching the literature, developing code, writing proposals to use telescopes, doing first-pass “reads” of their peers’ submitted proposals, and actually solving problems—to agentic AI systems such as Anthropic’s Claude or OpenAI’s Codex. Major institutions such as the Space Telescope Science Institute and the Institute for Advanced Study have rushed to hold meetings on LLMs and AI agents in science. In March, Anthropic published a blog post from Harvard University physicist Matthew Schwartz about his experiments in “vibe physics.” After supervising Anthropic’s Claude model closely enough to catch its many fabrications and bluffs, he forced it to generate in 2 weeks a real, publishable physics paper that he claimed would normally take a year. Schwartz’s sense was that the “AI grad student” approximated a second-year grad student at Harvard. Give AI 12 more months, Schwartz extrapolated, and LLMs’ capabilities may rival those of postdocs.

Representatives of the AI companies, who seem to view astrophysics problems as public relations boons that offer compelling showcases of their models, boast that their technologies will soon achieve supremacy over actual theoretical physicists, astrophysicists, and cosmologists. Some even make mechanizing the study of the night sky a selling point. In February, as Elon Musk began seeking to take SpaceX public to raise cash for orbiting data centers, his company explained its true goal as “scaling to make a sentient Sun to understand the Universe.”

Already, by making it faster and easier to produce professional-seeming papers, AIs threaten both to overwhelm journals and peer reviewers and to take opportunities away from junior scientists. But far upstream of that, many scientists interviewed by Science sense a phase change underway. Many fear that if unleashed in all parts of the scientific process, AI tools could lead to nothing less than the death of astrophysics as a human endeavor. “A lot of people think that it’s too late to intervene—we’re done,” says David Hogg, a computational astrophysicist at New York University (NYU).

Although fears of a robot takeover are inspiring soul-searching across society in general, astrophysics is a strange, special case. It relies on reservoirs of prestige, deep public appeal, and invocations of wonder: Human beings and funding agencies alike just seem to like studying the stars, and astronomical progress has long been thought of as a synecdoche for human progress in general. Yet because astrophysics is already mostly data science and math, many of its juiciest problems may be low-hanging fruit for LLMs.

At the same time, understanding the night sky promises little in the way of economic applications or human lives saved, so progress at all costs doesn’t seem as imperative as it might for fields such as drug discovery. If any scientific discipline could stake out a nuanced, human-centric relationship with powerful new AI tools, it would be astrophysics. So how is that going so far?

Early signs are, in a word, mixed. In September 2025, a guest speaker at the NYU physics department ran an AI agent in real time in the background. As he spoke, the system—called Denario and built by a group at the Flatiron Institute, a privately funded research center dedicated to advancing science using computational methods—generated entire scientific projects. It scoured journals, spun out ideas, carried out analyses, and extruded professional-seeming scientific papers—some obviously goofy, some plausible—that popped up on the screen behind him. With tools like this and beyond, he said to an audience of mostly grad students, you no longer need grad students. Why wait months for a young human scientist to do a project when an AI can give you the answer within an hour?

“It was a very bizarre talk,” says Matthew Daunt, a seventh-year grad student at NYU, because even Denario’s most impressive outputs didn’t look all that scientifically useful. “And yeah, the comment of ‘You don’t need grad students anymore’ kind of pissed me off.”

Hogg, who also works at the Flatiron, had a one-on-one meeting with Daunt right after the lecture. “He was like, ‘I’m not cattle,’” Hogg says.

The incident inspired Hogg to think about the moral value of AI systems studying the universe—and the value of the humans that such systems may replace. Simply banning the use of AI models would require an impossible amount of policing, he felt. But if astrophysicists simply sit back and let these models “cook,” their hundred-thousand–fold advantage in speed would mean scientists would be quickly buried under an unreadable avalanche of machine-only studies.

It seemed to him that finding a middle path between these extremes, in which LLMs help science and scientists alike, involved first wrestling with a question that seems less in the realm of physics and more like metaphysics. In February, he posted his paper to a preprint server . Its title: “Why do we do astrophysics?”

Hogg offered an answer. It wasn’t to solve the cosmos but to grapple with it: a journey, not a destination. Graduate students aren’t supposed to be only the means of the science—doing work for senior scientists—but an end in themselves, molded into competent scientists by doing that work. In so doing, today’s students would become the latest link in an unbroken chain of practice, going all the way back to the first among us who looked up agog at a sky full of stars. “Anyone working in astrophysics,” Hogg wrote, “is someone who wants to do astrophysics , not someone who wants to learn the answers .”

There are only a few thousand astrophysicists worldwide. Perhaps 100 email responses came in within a few days after Hogg posted the paper. Some correspondents disagreed. Many thanked him for dragging the conversation from coffee shops into the literature. “Flatiron fell apart,” Hogg says. In frenzied conversations, many of his own colleagues were compressing their internal timelines for when AI might bring major changes to their science from years to weeks—and fretting about the consequences.

Hogg had suggested I could visit the Flatiron to hear some of these conversations in person. But after weeks of emailing, a member of the Flatiron’s communications team turned me down. The institutional politics around AI and systems such as Denario, they told me on background, had become so fraught that my request stirred up a hornet’s nest of internal conflicts.

Astronomy is no stranger to advances in quantitative methods, having arguably given birth to them in the first place. Thousands of years ago, scholarly orders from Mesopotamia to Mesoamerica to China and beyond served as intermediaries between their societies and the cosmos by combining extensive bookkeeping with clever mathematical algorithms—what might now be called data science. In the second century B.C.E., Hipparchus accessed and reprocessed centuries of Babylonian records to develop theories of solar and lunar motions. At the beginning of the 1600s C.E., Johannes Kepler used Tycho Brahe’s observational data to fit mathematical laws to the motions of the planets.

As the supply of astronomical data exponentially increased, astronomers and mathematicians pioneered many of the first “computers”: human beings scribbling on paper. One early program employed teams of hairdressers put out of work by the execution of the aristocracy in the French Revolution to calculate logarithms. By the 1820s, mathematician Charles Babbage, dreaming of automating mental work the way the steam engine had automated physical labor, designed the first programmable computing engine to make astronomical calculations.

By the end of the 19th century, photography had began to be applied to the study of the night sky, leading to another explosion in data and computing needs and rendering a human eye on a telescope eyepiece irrelevant. Photonegatives of distant galaxies piled up in basements in places such as the Harvard Observatory. More scientific data could be extracted, astronomers understood, if only these images could be studied in detail.

In the Harvard Plate Stacks, three levels of hulking black cabinets store some 600,000 thin glass plates from this era. To process all these data from the cosmos, curator Thomas Burns explained, Harvard astronomers brought in a new workforce of woman computers. The women got a chance to make heroic contributions to a field that would otherwise have excluded them. But they were also exploited. Some, from working-class backgrounds, needed a paycheck too much to negotiate. Others, from wealthy families, didn’t need the paycheck at all.

Burns descended a narrow spiral staircase and fetched a glass negative from 1934. Photons from thousands of galaxies around the Coma Cluster had traveled 300 million light-years to Earth, where a chemical reaction in the film had trapped their energy like fossils in amber. The glass, illuminated against a white backlight, was covered in tiny markings in different colors. First a “computer” named Muriel Mussells Seyfert had labeled countless galaxies. Then others had revisited the same frame in other passes, leaving lasting traces of a layered dialog.

Burns explained that Seyfert, uncredited, had helped calculate the mass enclosed within a red circle around the center of the cluster. The result corroborated an odd contemporary proposal that the galaxy cluster harbored some extra, unaccounted-for “dark matter.” Another astronomer, Vera Rubin, came across these same underappreciated data in the early 1950s. Perhaps inspired by this hint, she would later go on to find even firmer evidence for the existence of dark matter.

Another three-quarters of a century later, in March, NVIDIA announced its new Space-1 Vera Rubin Module , designed for use in swarms of orbiting data centers. Their proliferation could pollute astronomical images—and the AI models they enable could render astrophysics itself unrecognizable. I asked Garraffo about the scenario in which her field produces fewer human stories like the one captured in this annotated glass plate. What if people—such as Seyfert, or Rubin—no longer make these kinds of discoveries themselves?

“What I love is to chase the truth,” Garraffo replied, considering. Perhaps the moment of human insight frozen on the plate was romantic, but the work had clearly also been tedious—and to her, an elegant truth of the universe such as general relativity was more romantic. If tools exist that would lead to such truths faster and better, she argued, astronomers should use them. “Pretending, ‘Let’s cover the truth so people feel like they can discover it’—it’s very narcissistic, in a way.”

Garraffo isn’t sure when AIs might surpass human capabilities, but a recent experience convinced her they haven’t yet. In her early career, as a pen-and-paper theoretical physicist, she had worked on alternate versions of Albert Einstein’s theory of gravity, general relativity, which is known to be incomplete. Garraffo wanted to solve the equations for one alternative, called Einstein-Gauss-Bonnet, and find an exact description within the theory of the shape of spacetime around a rotating black hole. It was a practice case, she thought: If you could figure out how to do that, the same conceptual tools might work across other alternate theories of gravity, giving scientists a new way to probe and evaluate new theories that aim to usurp Einstein’s. She trained a neural network to intuit its way to an approximate, numerical solution. But this kind of network was not an LLM, and it couldn’t rewrite its inner workings out into a neat, concise mathematical answer.

Then she asked the best publicly available versions of both Claude and ChatGPT to solve the Einstein-Gauss-Bonnet equations for a rotating black hole analytically and produce the kind of crisp answer she sought. It was a very hard but well-posed mathematical problem, and she was optimistic it could work. “The models I tried failed miserably. Miserably,” she said. Claude, in back-and-forth exchanges, claimed it could find an interesting result for the problem, then wrote up an entire impressive-looking paper. But as she read the paper, she realized it was just stating existing ideas in byzantine ways.

If the LLMs had been able to make progress on the problem, and if AI systems stopped improving right at that level, “that to me, maybe selfishly, would be ideal,” Garraffo said. With tools of that strength, humans like her would still need to identify the relevant questions, ask them, interpret the answers, and have the fun of making sense of what that meant for physics. “Then we have to press the brakes.”

But the power of the models may soon surpass that level. “We all collectively came to the realization that these tools are about to take over,” postdoc Rodrigo Córdova Rosado told me in his office earlier this spring, flanked by a LEGO X-Wing on top of a shelf of physics textbooks. In about 2 hours, he told me, he had used Claude to interpolate between the best existing general relativity textbooks and write a new one, complete with sample problems and figures generated by scripts it wrote in the math software Mathematica. He was now checking the book for errors. (On the side, as a student of the Osage language and a member of the Osage nation, he had been experimenting to see whether AI might be able to create an Osage-English dictionary.)

“This is a wave,” he said. “If you do not surf it, it feels like you’re gonna get drowned by it.”

During my visit , Garraffo invited team members who wanted to talk more philosophically about AI to come to her office. Seven scientists packed in, two more joined remotely, and they began an impromptu roundtable discussion.

The conversation grew loud. Do AI agents lack good “taste” in scientific problems? Were these tools democratizing, helping non-English speakers participate in science—or were they antidemocratic, by making science dependent on subscriptions to some of the wealthiest corporations in the world? Were astronomers now free to care less about computer science and more about the universe? Would the fact that LLMs are inherently probabilistic—and not unerring engines of mathematical logic—make the math and statistical analyses they spit out untrustworthy? “Is it insane,” one grad student exclaimed, “if I don’t see this as such a big issue?”

Both in that room and outside it, many of astrophysicists’ concerns lump into two families. First, many researchers fear LLMs’ ability to effortlessly conjure text, code, and highly technical analyses threatens the current metrics, incentives, and trust networks that knit “astrophysics” into a cohesive, reliable body of knowledge.

For example, in a publish-or-perish world, publishing papers at a fast rate and racking up citations has been treated as a proxy for the quality of a scientist’s output, which in turn determines who gets resources to do research at all. But now publishing an incremental but impressive-looking paper seems cheaper and easier than ever. “LLMs are forcing us to face the fact that, as a field, we do not do well at assessing ourselves and our peers,” Natalie Hogg, a cosmologist at the University of Cambridge, wrote in a February blog post .

Ethan Vishniac, editor-in-chief of journals published by the American Astronomical Society (AAS), is already dealing with this issue. Since LLMs first became popular, he says, paper submissions have surged, making it harder to find reviewers. Much of the increase from established astronomers has since leveled off, but paper submissions from outside the known community have continued to increase. In April, for example, a “20-year-old working alone from Chennai,” India, claimed in a blog post that an end-to-end science pipeline of his own creation had conceived of, executed, and written a study on a distant source of radio signals. That paper was accepted in March by an AAS journal but was later rejected for not disclosing the nature of its AI use, his post said.

Traditionally, AAS editors have responded to every submitter with personal feedback. But now, Vishniac says, the correspondence sometimes gets so “weird” that he suspects he’s talking to agents, not people. Rather than democratizing science, these tools may soon force journals in the opposite direction. “The quantity of things of low quality can strangle the system,” Vishniac says. “And the only solution to that is to do pretty much arbitrary gatekeeping.”

The second cluster of worries involves what researchers outside of astronomy have termed “deskilling” or, more bleakly, “cognitive surrender.” What if AI-dependent astrophysicists, especially young ones, lose or never build their own math, coding, and reasoning skills? “At that point, we’ve just completely selected for the disappearance of science in 50 years, because nobody will know how to do anything,” Córdova Rosado said.

What AI enthusiasts dismiss as now-optional “grunt work” is exactly where graduate students develop the skills and intuition to do high-level science, Minas Karamanis, a cosmology postdoc at the University of California, Berkeley, argued in a March blog post . “Every hour you spend confused is an hour you spend building the infrastructure inside your own head that will eventually let you do original work,” Karamanis wrote. Once a student crosses the hard-to-see line where the machine lets them skip thinking for themselves, “you haven’t saved time. You’ve forfeited the experience that the time was supposed to give you.”

None of the scientists interviewed for this piece seemed confident projecting whether, in a few months or a year, their concerns will seem embarrassingly overblown or already painfully out of date. But in a field already threatened in the United States by sharp cuts to science funding and threats to the visas of foreign-born students, the situation for young scientists is “nightmarish,” Vishniac told me. “Why is this happening in 2026?” Hogg said, rhetorically. “Can we postpone the AI revolution in astronomy?”

The issues are closest to home for the supposedly replaceable grad students. Poking my head into offices at the Center for Astrophysics, I offered to anonymize any grad student who wanted to participate in this article. Just one, a physicist, came forward. The student said their peers talk about these issues all the time. They’re aware of Schwartz’s “AI grad student” post, which, the student said, “implies that everyone sitting in that office is obsolete until they become a third-year grad student.” But they were skeptical they could be so easily replaced.

And of course they, too, were using LLMs and agents. The grad student said their peers “are trying to [be] realistic: ‘This is about to be our future, this is what our careers are going to be.’”

During the discussion in Garraffo’s office, one of the most humanist, tech-skeptical voices had been Rafael Martínez-Galarza, AstroAI’s deputy director. The next day, in his office, he showed me how he was using Claude to contribute to a code repository and to build a machine learning model tailored to solar physics. He had learned how to code and supervise students before AI arrived, he said, putting him in a “sweet spot” to profit from these tools without blindly trusting them.

“I didn’t want to say this yesterday in the room because there were students there,” he volunteered, “but I can see how a lot of things that I sit here to discuss with the students, I can just transform into a prompt and get the results much faster. But do I want that?”

External pressures to be more productive might push him toward an LLM, he admitted. And he did share some of the stated goals of AI engineers who want their models applied to astrophysics. He, too, deeply wanted to understand the universe and see humans exploring the Solar System. “I suspect a lot of why astronomy and cosmology are so catchy is because at the very end, it is really about meaning, right?” he said. “I see value in the process of matter turning into neurons trying to understand itself. I think it’s beautiful—it’s almost poetic.”

But he conceives of that process, science, as a social activity, by and for humans. It would be more rewarding, he thought, to spend months solving a problem with a human than minutes with even the most brilliant chatbot. Even if answers to the cosmos could spill out in a sudden, oracular revelation, why not delight in getting to ask the universe our own questions and working through them together?

“I don’t see why we would rush colonizing the Galaxy or even understanding the universe, if that means a step back in the human experience,” he added. “I don’t see the point.”

The top Democrats on a Republican-led congressional committee that historically opposes most U.S. interactions with China are slamming President Donald Trump for firing all the members of the board that oversees the National Science Foundation (NSF).

Until now, Democrats on the U.S. House of Representatives Select Committee on the Chinese Communist Party (CCP) have largely fallen in line with Republicans on the panel, chaired by Representative John Moolenaar (R–MI). But today’s letter to acting NSF Director Brian Stone suggests that alliance is fraying, and that some Democratic members are putting limits on supporting how Republicans are addressing the growing challenge China poses to U.S. leadership in science, technology, and the global economy.

“These firings are another example of how the Trump White House is harming our competition with China by carelessly cutting budgets, disempowering key presidential appointees, and undermining scientific merit review,” Representatives Ro Khanna (D–CA) and Haley Stevens (D–MI) write in the letter to Stone , who has led NSF on an interim basis since Sethuraman Panchanathan resigned as director in April 2025.

The letter asks Stone to explain Trump’s 24 April decision to remove all 22 members of the National Science Board (NSB). The two lawmakers also want NSF to release an analysis of how China and the United States stack up scientifically that was written by NSB but never officially issued. The two-page statement highlights a “trend that’s been clear for a few decades and across several administrations,” says former board member and physicist Julia Phillips, adding that “it’s an important message that needs to be conveyed to the public.”

The select committee was created in January 2023 after Republicans regained control of the House. Moolenaar has not publicly commented on Trump’s decision to disband NSB. But he has previously won support from committee Democrats in chastising U.S. agencies and universities for inadequately managing joint research efforts with Chinese researchers and institutions, saying those collaborations have allowed the CCP to steal U.S.-funded innovations.

More recently, however, Democrats have so far declined to co-sponsor legislation introduced by Moolenaar that would prohibit federally funded U.S. scientists from interacting with anyone “affiliated with” a Chinese institution on one of the U.S. government’s many blacklists. The bill’s sweeping definition of collaboration includes co-authorship, sharing data, material transfers, and any joint supervision of students, and the blacklists cover any Chinese university, laboratory, or hospital contributing to the country’s “military-civilian fusion strategy.”

Today’s letter from the two Democrats criticizes Trump’s proposed budget cuts to NSF and other science agencies, his administration’s attempts to prevent agencies from spending money already appropriated by Congress, and personnel decisions it alleges “have interfered with civil servants doing their job.” Citing coverage in Science of the board’s unreleased statement on how China now outspends the U.S. on research, it asserts that “Trump allowing China to surpass the U.S. in scientific R&D is the final straw.”

The letter asks Stone “to identify who ordered the summary dismissal of the Board [members] and their rationale for doing so” as well as NSF’s response to the board’s analysis of the status of U.S.-Chinese competition. An NSF spokesperson said the agency will respond privately to the letter, which was also sent to the White House office that informed board members they were being canned.

The U.S. government has charged two scientists at the National Institutes of Health (NIH) with criminally conspiring to smuggle biological samples into the country and lying about it. FBI testing of some of the confiscated test tubes the pair was ferrying from Africa revealed they contained monkeypox viruses, but ones that had been inactivated, rendering them unable to infect people. The samples may have been used for routine diagnostics, a virologist has told Science .

Vincent Munster, who heads the virus ecology section at a division of the National Institute of Allergy and Infectious Diseases (NIAID), and his research fellow Claude Kwe nonetheless face a maximum sentence of 5 years in prison for allegedly not properly declaring to U.S. customs officials in January the contents of their luggage after returning from a trip to the Republic of the Congo. They had been assisting investigators dealing with an outbreak of monkeypox virus, which causes the disease now known as mpox—symptoms are often limited to a painful rash, but it can be deadly. According to the complaint, the two researchers, who work at NIAID’s Rocky Mountain Laboratories, told customs authorities the tubes carried diagnostic reagents.

Munster and Kwe surrendered themselves yesterday to authorities and were released on their own recognizance. They pleaded not guilty today at a brief court hearing in Montana.

The criminal complaint against them was first filed on 17 March, but only unsealed yesterday. It comes amid a recent U.S. crackdown on potential violations by researchers of the rules and regulations regarding the import of biological materials. Earlier cases focused on Chinese nationals working as scientists in the United States, resulting in multiple firings and deportations —and protests that President Donald Trump’s administration was criminalizing relatively minor infractions for a political agenda. One case was prosecuted by the same U.S. attorney, James Gorgon Jr., whose office brought the complaint against Munster and Kwe.

A press release from the Department of Justice notes in its headline that the two NIH scientists are “foreign nationals.” Munster is a citizen of the Netherlands and Kwe a citizen of Cameroon. “These NIH experts apparently broke our laws by smuggling viral pathogens on a packed commercial airplane from an outbreak in the Republic of Congo,” Gorgon said in the release. “Let that sink in.”

Several virologists in the U.S. who bring pathogen samples into the country would only speak with Science off the record about the case because of fear of repercussions if identified. “Clearly, the Justice Department is going way over the top because there’s a political point to be made,” one said. Another, though, called it “ill-advised” if indeed Munster and Kwe used personal luggage to carry viral samples, rather than sending them via approved shipping containers.

Marion Koopmans, a virologist who works at Erasmus Medical Center, where Munster completed his Ph.D., says she hopes this is not part of a “political witch hunt” by the U.S. government, which she believes has recently targeted other infectious disease researchers. “This is serious, but it is hard to see this as anything other than an honest mistake,” Koopmans says.

And the inactivated viruses in the samples may actually have been for diagnosing diseases, says Angela Rasmussen, a virologist at the University of Saskatchewan who develops diagnostics. “Inactivated monkeypox viruses routinely are used as a control in diagnostic tests or to develop the assays,” she says, noting that Munster’s lab has published on this.

A spokesperson for the Department of Health and Human Services (HHS), NIH’s overseer, said, “NIH is committed to maintaining the highest standards of biosafety, biosecurity, and stewardship of research materials.” They emphasized that “there was no risk at any time to the staff or public in or around” the Montana lab.

Munster and Kwe could not be reached for comment.

Journalist Paul Thacker, a former Republican Senate staffer, last month first reported in his online newsletter that he had seen HHS emails saying Munster and Kwe (who also uses the last name Yinda) were under investigation for smuggling monkeypox viruses, although he misidentified the country from which they were returning. Thacker has long sounded the alarm about what he regards as risky virus studies, contending along with others that U.S. researchers helped spark the COVID-19 pandemic by supporting research on a coronavirus that then leaked from a lab in Wuhan, China. On 26 May, Senator Tim Sheehy (R–MT) cited the Munster and Kwe allegations in a letter to HHS’s inspector general, calling for an investigation of Rocky Mountain’s safety practices. “We don’t want Montana to be the next Wuhan,” Sheehy wrote in a social media post that included his letter.

According to the complaint, Munster and Kwe flew into the Detroit Metro Airport on 25 January after 9 days in the Republic of the Congo. Customs and Border Protection selected Kwe for secondary inspection and asked him to retrieve his luggage, noting “nervous behavior.” Kwe put a large black plastic case—“atypical of business travel”—near Munster, so both were detained. Munster allegedly said the case contained “diagnostics and testing equipment” and that he had the required documentation for importing the samples on his laptop, but that they wouldn’t be needed. “I do this all the time,” Munster allegedly said, and he “adamantly denied” that the case had biological samples.

The customs officers sent the scientists to an agricultural specialist to inspect the case. Munster then went on his laptop and provided the requested documentation, the complaint states. Opening the case, the specialist found two plastic containers with 113 microcentrifuge tubes in Styrofoam coolers. An FBI laboratory later tested 20 of the samples and found 17 had DNA from monkeypox, one from chickenpox virus, and one from human DNA. “A set of samples tested did not propagate and thus are assessed to be inactivated,” the complaint says.

U.S. importation of inactivated viruses has “more permissive requirements,” the complaint notes, but it faults Munster and Kwe for not presenting the “true identities” of the contents in their luggage or having the proper certifications. The complaint says they also violated NIH regulations by carrying noninfectious biological material on a commercial aircraft without a permit.

Munster is a widely respected virologist and has won several prestigious awards for his work. He has studied many lethal viruses, including Nipah, Ebola—for which he tried to find the natural reservoir in the Republic of the Congo—and the coronaviruses that cause Middle East respiratory syndrome and COVID-19. He has published more than 400 scientific articles, 12 with Kwe.

Kristian Andersen, an evolutionary biologist at Scripps Research who studies several of the same pathogens, says he believes the criminal complaint makes unsupported arguments about importation regulations. “The offenses, if any, appear extremely minor and much of this appears to be based on basic misunderstandings of how science is conducted,” says Andersen, who himself has been subjected to intensive scrutiny for his work that supports the theory COVID-19 began when a virus naturally jumped from animals sold at a Wuhan market to humans.

Munster and Kwe will have a court appearance on 24 June in Michigan.

For decades, string theory promised a “theory of everything” that described all particles and forces as tiny vibrating strings. Physicists hoped it could also solve one of the field’s deepest problems: reconciling quantum mechanics with gravity. But as string theory grew increasingly elaborate—and experimentally unreachable—many physicists lost hope.

Now, some researchers are revisiting the theory from first principles. In a paper in press at Physical Review Letters , Clifford Cheung, a physicist at the California Institute of Technology, and colleagues lay out a small set of assumptions about the universe and show that they inevitably give rise to string theory. The work is part of a broader revival of the “bootstrap” philosophy that inspired string theory in the first place: building up explanations from a small set of consistent, general principles rather than deriving laws from a particular mechanistic framework. “It’s a trend away from a specific model that someone came down from the mountain with on two tablets,” Cheung says. “People are kind of going back to the basics.”

The approach does not prove string theory is correct. But, “It’s quite remarkable that with fairly minimal assumptions you are led to string theory,” says Andrew Tolley, a physicist at Imperial College London who works on an alternative model of gravity . Clarifying what assumptions underlie the theory could help limit the range of possible versions of string theory and its competitors, he adds. “It’s tremendously interesting to know what is allowed or not.”

String theory traces back to the 1960s, when Italian physicist Gabriele Veneziano, using only particle collider data and basic principles about nature, discovered a formula that seemed to capture the behavior of elementary particles. Although physicists soon found that standard quantum theory could explain particle behavior in a more systematic way, Veneziano’s equation gained a second life when researchers realized it described particles not as points, but as vibrating strings.

Over the following decades, string theory grew into an ambitious framework in which all the fundamental particles and forces arise from tiny vibrating strings, much as different notes emerge from guitar strings of different tensions. Because strings are extended objects rather than infinitesimal points, the theory sidestepped mathematical headaches that plagued other attempts to merge quantum theory with gravity. The dream was essentially that “you write down a single magic formula, and somehow everything else follows from this,” says Renate Loll, a theoretical physicist at Radboud University.

But the theory grew ever more baroque. For it to work, strings would need to exist in at least 10 dimensions—the four familiar ones of space and time plus six additional dimensions curled up beyond perception. The theory also required certain symmetries not seen in nature and permitted an unimaginably vast number of possible universes. Furthermore, the strings are too small to ever probe directly, making it difficult to falsify. For Peter Woit, a Columbia University mathematician and hardened critic , string theory is a vague and untestable idea blindly adopted with a tribal allegiance. “It’s just not science—it’s a really insane situation,” he says.

Cheung, who does not identify as a string theorist, agrees that “there’s been a gap between the promise and the execution,” he says. “It has not given us a theory of the world we live in, which was its entire point.” But he argues that this doesn’t make string theory worthless. Even Woit acknowledges that string theory has generated fruitful ideas across physics and math. Today, theoretical physicists are learning to use string theory as a helpful framework that doesn’t have to live up to its original promises.

To get a better handle on the theory, they’re stripping it to its essentials. Cheung’s study, along with another one posted to arXiv in January, starts with two reasonably conservative assumptions: that the probabilities of all possible outcomes of an event add up to 100%, and that the laws of physics are consistent for observers moving at different speeds. Each group then posits additional assumptions that have not been borne out by observations. Cheung’s analysis invokes “ ultrasoftness ,” the idea that the probability of certain particle interactions drops off at a particular rate at high energies. The second study, led by University of Michigan physicist Henriette Elvang, instead assumes “supersymmetry,” a maximal coupling between matter and forces. Both groups conclude the only theory that can satisfy their assumptions is one that looks like string theory.

Woit accuses both efforts of circular reasoning, as there’s no experimental evidence to justify either of these special assumptions “unless you want to get string theory to come out of it,” he says. But Cheung and Elvang stress that their aim is not to prove the inevitability of string theory. “I don’t have a dog in the fight; I just work here,” Cheung says. Rather, the goal is to explore the space of possible theories under rigid constraints—regardless of whether they reflect reality. “It’s not the real world, but it’s a toy model that allows me to build my tools and have control,” Elvang says. “Then we can let go of the assumptions and build up from that.”

Loll welcomes the “backlash back to first principles” but questions whether these string-inspired assumptions are the right starting point to build up from, given how little the theory seems to resemble our world. “We’ve been thinking completely outside any boxes for too long,” she says. She believes the quantum gravity field should ground itself by finding reality checks in experiments or numerical simulations. “We’re all very far from that, and it’s a dangerous territory,” she says.

But Tolley thinks the bootstrapping effort itself could inspire new experimental probes. He has applied the same strategy to help constrain different cosmological models, for example describing how the universe expanded over time. Because some of the bootstrapping assumptions tie together the behavior of particles at different energies, “I can use low-energy physics to infer something about high energies,” he says. He suggests the same relationship could yield low-energy predictions for string theory that could be tested experimentally.

The one thing the researchers all agree on is that the field would benefit from more alternative models to string theory. Cheung sees the agnostic, bottom-up exploration as a step in that direction. “You can either give up on the problem because it’s too culturally toxic, or you can ask: If you want to find an alternative, what do you need?” he says. “Now, we know exactly what to do.”

As artificial intelligence (AI) muscles its way into field after field, researchers have wrestled with what it means for the future of their disciplines. Few communities have felt that pressure more acutely than mathematicians, who have haplessly watched AI get frighteningly smart, frighteningly fast.

Today, 16 math specialists have turned that unease into a public cry for help—and call to action. Part warning, part manifesto, the 11-page Leiden Declaration on Artificial Intelligence and Mathematics cautions that unchecked automation threatens not only how math is practiced, but what the discipline stands for. It also lays out principles for using AI in ways that support, rather than erode, the field.

For years, AI researchers have used math as a proving ground for their models . By their nature, math problems are easily gradable and offer an effectively unlimited training pool, and building AI models that can prove theorems is widely seen as a steppingstone toward higher performing reasoning systems. But AI is no longer confined to just straightforward tricks. Last month, OpenAI reported one of its newest general-purpose reasoning models had independently solved a famous, 80-year-old conjecture —only the latest feat in a series of breakthroughs pushing AI to the frontiers of the subject.

However, the declaration argues math is more than a machine for producing correct answers. The discipline, its authors believe, is a deeply human endeavor built on creativity, understanding, collaboration, and the pursuit of knowledge for its own sake. Those values often clash with the incentives driving AI development. “The tech industry proceeds in accordance with commercial logic, which is antithetical to the values of mathematics,” declaration co-author Michael Harris of Columbia University told The New York Times .

The authors warn the consequences are already becoming visible. AI-generated papers could overwhelm peer-review systems with low-quality work , make it difficult to assign proper credit for discoveries, and disadvantage researchers who choose not to rely on AI tools. The authors also raise ethical concerns about mathematicians’ work being used to train AI systems for military and surveillance purposes.

The declaration, which is endorsed by the International Mathematical Union, the discipline’s leading global body, is now open for signatures from individuals and organizations worldwide . It will also be discussed at next month’s International Congress of Mathematicians in Philadelphia, which comes as governments are also beginning to take a greater interest in the advancement of AI systems. Today, for example, President Donald Trump signed an executive order aimed at giving federal officials advance access to the most capable AI models before their public release , in part so agencies can prepare for any security threats AI might enable.

Even in a hype-drenched field, the claim seems bold: Tech giant Microsoft announced today it plans to build a useful quantum computer in just 3 years—half the time it had previously cited. The schedule can accelerate, Microsoft researchers say, because of improvements to the building block of their technology: a microscopic device called a topological qubit.

“We’ve talked about a scalable quantum machine on the 2033 timescale, and we believe that these kinds of accelerations can bring that into 2029,” Chetan Nayak, the physicist who leads Microsoft’s quantum computing effort, said in an online press conference last week. But critics maintain that the paper reporting the improved qubit lacks evidence that the device actually works. “The Microsoft Quantum project follows a sustained pattern of unreliable claims, so the new ones are not surprising,” says one of the most vociferous doubters, Sergey Frolov, a physicist at the University of Pittsburgh.

A quantum computer would exploit strange quantum phenomena to solve certain problems that would overwhelm any conventional supercomputer. Whereas a conventional computer employs bits that can be set to 0 or 1, a quantum computer manipulates qubits that can be set to 0 and 1 simultaneously. Thanks to those superposition states, a quantum computer can analyze many inputs at once, with the wavelike quantum states interfering to cancel out wrong solutions and cause the right one to pop out.

In principle, a qubit can be any quantum system with two states to denote 0 and 1. Scientists have fashioned qubits out of tiny circuits of superconducting metal resonating with unquenchable currents, individual atoms , lone ions , and single photons . Most qubits’ delicate quantum states quickly fuzz out when exposed to noise from the environment. That’s one reason why today’s quantum computers contain at most a few hundred qubits.

To obtain more-robust qubits, Microsoft researchers have been pursuing a particular byzantine design. On a chip made up of semiconducting layers, they lay down a strip of superconducting metal, which induces a wirelike region of superconductivity in the underlying semiconductor. Electrons pair up in this region, and a tiny electrode called a quantum dot can inject an additional, unpaired electron. “Within one of these wires we’ve got an even or odd number of electrons, maybe 10 million versus 10 million and one,” Nayak says. The state with no lone electron signifies 0, and the state with one signifies 1. Microsoft researchers read out the states by measuring the wire’s capacitance with another nearby quantum dot.

In theory, quantum effects should make those states particularly robust. Weirdly, the lone electron is delocalized so it exists at every location along the wire at once, making it harder to disturb. In addition, the unusual shape of the quantum state describing a pair of electrons reduces the chances a pair will interact with its environment—the topological aspect of the device.

Reports of such delocalized electron states—which are known as Majorana zero modes —have come and gone over the years. In February 2025, Microsoft researchers reported, to intense controversy, that they had achieved Majorana states in devices fashioned by laying down traces of superconducting aluminum on the microchip. Those states persisted for about 2 milliseconds before flipping, the researchers claimed. Now, they have replaced aluminum with lead to increase the strength of the binding of the pairs in the semiconductor, and the key quantum states last 20 seconds, Nayak says. That’s long enough to put a full-fledged quantum computer within reach, he says.

Critics aren’t buying it. Achieving the Majorana states requires applying a strong magnetic field along the wirelike region and precisely tuning a voltage that controls the density of electrons in it. Microsoft researchers have not shown they can tune into the rare, desired state, says Henry Legg, a physicist at the University of St. Andrews. Rather, he says, the data suggest the system is settling into a quiet spot in a landscape of noise.

Moreover, he says, there is a simpler explanation for the switching behavior the Microsoft team is seeing. To control and read out a nanowire, they use a number of quantum dots surrounding it. The switching behavior they see could just be an electron hopping on and off a quantum dot, perhaps one formed incidentally by part of the wirelike region, Legg says. “This is exactly what you could get from a quantum dot.”

Even assuming the Microsoft researchers have achieved the Majorana states, there’s no evidence their device actually works as a qubit, Legg and Frolov say. For a practical qubit, researchers need to control the device and ease it into different admixtures of its two states. But the new paper contains no such demonstration. Legg also notes that the endurance time Microsoft quotes does not necessarily equal how long a superposition will last.

Nayak acknowledges that Microsoft researchers have been cagey with their peers in the past. “We said just trust us, we’re going to make the topological phase more robust,” he says. “Now, we’re revealing more about how we’ve been doing that.” He says Microsoft researchers also have data, albeit unpublished, that show they can fully control their qubits and can even run quantum algorithms on their chip.

That “trust us” approach rankles other scientists. “You go to a conference and somebody mentions Microsoft and people are chuckling,” Frolov says. “It’s become a joke and it’s terrible for the field.” To achieve a scalable quantum computer in 3 years, Microsoft researchers would likely need to have a prototype running in the lab now, he says. Based on the results they published so far, that scenario is “implausible,” Frolov says.

Legg thinks the Microsoft researchers have convinced themselves of conclusions their data do not support. “At this point, I think that this is less the realm of reality and headed to something more like a quasi-religion.” By Microsoft researchers’ own scheduling, judgment day will arrive quite soon.

The European Union’s push to attract foreign scientific talent took a step forward today, with applications opening for a new supergrant aimed at “outstanding researchers” from around the world. Part of the EU’s Choose Europe for Science initiative launched last year , the 7-year, €7 million grants will be the largest ever provided by the European Research Council (ERC), which funds basic, curiosity-driven research. The awards could prove a powerful draw for U.S. scientists disillusioned with the state of research at home.

The new grants—dubbed ERC Plus—come as European science faces a transition. The EU’s €95.5 billion research funding programme, Horizon Europe, has 18 months left to run and the European Parliament and member states are now discussing the shape of its successor . ERC could end up with a significantly larger budget—but other ideas, such as reducing the length of the ERC president’s term, have stoked debate about the council’s independence.

The current ERC president, Maria Leptin, spoke with Science about the new grant program and the agency’s future. The interview has been edited for brevity and clarity.

Q: Why does the world need ERC Plus Grants now?

A: What the world needs is peace and sanity. We are talking about what is good for the research community in Europe. We did it for two main reasons. We are doing something that demonstrates the EU is attractive to high-level research careers. We’ve [also] been talking about more generous awards. For high-risk, big bold ideas, 5 years [the maximum length of existing individual ERC awards] is often not enough. Other funders have recognized that and offered longer terms.

Q: ERC will make just 30 ERC Plus Grant awards. Some scientists have estimated success rates for the scheme could be as low as 0.5% . They argue the process could be wasteful. What do you say to that?

A: I don’t talk about rates because they are pretty meaningless. The rate depends on the number of applicants. There is a lot of gaming, guessing, and self-selecting in the community. [A success rate of] 0.5% is nonsense as we are not going to have everybody who applies for a normal grant applying [for an ERC Plus Grant].

Q: The ERC Plus Grants are open to researchers at any stage of their career. Given that prior research success is a prerequisite for applying, how are you going to ensure more junior researchers are not excluded? Will you use quotas or some other measures?

A: We are not going to have quotas, we are going to look at the best proposals. The panels are extremely experienced. When we talked in the Scientific Council [of ERC], they agreed that panels will recognize people for who they are at their [career] stage. We are all used to doing that, for prizes, faculty recruitment, and promotion. There was also a consensus that, on the whole, the more fun ideas, the more forward-looking ideas … come from younger people.

Q: The European Commission has proposed doubling ERC’s budget in the next iteration of Horizon Europe. What would you do with that money?

A: The Scientific Council are discussing that at the moment. A doubling of the budget will already provide restraint because if we wanted to inflation adjust the grants that would require a 1.5 times [budget] increase. We also can’t fund all the projects that are ranked excellent [at the moment], we only fund 60%. Funding the other 40% would be another more than 1.5 times [increase]. If ERC Plus Grant turns out to be interesting, we would also like to continue it. [Current funding for the new grants comes from an additional pot of money provided by the Commission.] We have many ideas about smaller exploratory grants. So we will be spending this year dreaming.

Q: Last year, ERC’s Scientific Council wrote to the Commission asking for ERC to be established as an independent body, with a stable long-term budget. Why do you think this is necessary?

A: We wanted to recall that when ERC was founded, the aim was to create an entity that mirrored independent national funding agencies that are at arm’s length of government. Obviously, there is huge support for ERC within [Horizon Europe], but you only have to look outside to see what can happen if an independent research program is not fully protected politically. Who knows what the governments of Europe will look like in 7 years.

Q: You’ve publicly opposed the Commission’s plan to decrease the term of ERC presidents from 4 to 2 years. Why would that be a bad thing?

A: After 2 years in this job, I [had] only just learned the ropes. You just need a little bit of runway to plan and implement these things. No other organization of this type have executives with 2-year terms.

Q: Earlier this year, ERC announced it would require unsuccessful applicants to wait an extra year before resubmitting grant proposals, because it was being overwhelmed with submissions. This outraged researchers and you quickly rowed back . How are you now managing the large number of applications?

A: We’ve introduced a number of small tweaks that we hope will make it easier for [review] panels. Less feedback for applicants. The community appreciates the feedback, so it’s painful. Fewer readers of each application. We don’t have a magic wand, it will take some work to find a way to streamline the process. All funders and journals are seeing similar increases and it is completely coincident with the rise of AI [artificial intelligence]. And, of course AI helps to write, so it’s hardly a surprise.

Q: Will you consider using AI in the evaluation of proposals?

A: We will look at using AI in the evaluation process. Now, we’re all scientists ourselves, so we don’t do these things without proper experiments with proper controls. We’re planning a bunch of experimental scenarios, and we will obviously not do this without consulting experts in the field.

When violent eruptions from the Sun slam into Earth’s magnetic field, they do more than paint aurorae across the night sky. They can scramble the electronics of satellites and induce powerful ground currents that knock out electrical grids. It’s been estimated that a one-in-a-100-year solar storm like the 1859 Carrington Event could cause more than $3 trillion of damage to the power grid alone.

Yet for decades, society’s only defenses have been better space weather forecasts and more durable technology on the ground and in space. Now, a small group of space physicists says humanity should intervene and weaken solar storms in real time. In a study published today in Space Weather , the researchers describe a provocative proposal called “StormWall”: a fleet of satellites that would release hundreds of tons of gases into space just before a solar storm strikes Earth. Computer simulations suggest the artificial cloud could cut the intensity of a major solar storm by half or more. “It’s as if you could install an airbag in the magnetosphere,” says Daniel Welling, a co-author and space physicist at the University of Michigan.

Call it “helioengineering”—a deliberate intervention in the near-Earth space environment. But unlike controversial geoengineering proposals to mitigate global warming, which would inject long-lived Sun-blocking particles into the atmosphere , StormWall’s protective gases would dissipate within hours, says Brian Walsh, the study’s lead author and a space physicist at Boston University. “It’s waiting for us to do some temporary modification.”

The proposal would require more extensive simulations and testing. But it is “highly innovative and appears to be quite feasible in the near term,” says Allison Jaynes, a space physicist at the University of Iowa. It’s a “laudable idea,” adds Gurudas Ganguli, a space physicist at the U.S. Naval Research Laboratory (NRL).

Normally, Earth’s magnetic field acts as a shield, deflecting most of the solar wind—the stream of charged particles that constantly flows out from the Sun. But during extreme solar storms, Earth’s magnetic field can “reconnect” to the Sun’s, short-circuiting the barrier. “It’s sort of like a shield dropping,” says Amy Keesee, a magnetospheric physicist at the University of New Hampshire.

When that happens, energy and charged particles stream into Earth’s magnetic environment. The charged particles can fry satellite electronics and harm astronauts . Some of the energy gets dumped into the upper atmosphere, heating and inflating it. The resulting drag can pull satellites out of orbit more quickly.

At ground level, the disturbances to Earth’s magnetic field generate electrical currents along any long conductive pathway . Those currents surge through power lines and transformers, creating blackouts such as the collapse of Hydro-Québec’s grid in Canada during a massive 1989 storm.

Earth has a partial natural defense. Just outside the planet’s atmosphere is a doughnut-shaped reservoir of cool plasma, or ionized gases. When a solar storm disturbs Earth’s magnetic field, a plume of this material is pumped toward the Sun. Observations show this extra mass acts like ballast, slowing the rate at which Earth’s magnetic field reconnects with the Sun’s. “It’s like a dead weight,” says David Sibeck, chief of heliophysics at NASA’s Goddard Space Flight Center.

StormWall would boost this natural phenomenon. In the researchers’ preliminary proposal, six heavy-lift rockets, such as SpaceX’s Starship or China’s Long March 9, would launch spacecraft the size of a small bus to geosynchronous orbit, roughly 36,000 kilometers above Earth—far beyond the International Space Station but still inside Earth’s magnetic shield. The spacecraft would carry giant canisters of lithium, barium, or sodium gases—or even just saltwater. When satellite monitors detect a hazardous solar eruption, the canisters would dump their contents. Sunlight would rapidly ionize the material into a plasma that would be drawn toward the dayside of Earth and hamper reconnection. The researchers estimate 400 tons of material would reduce the strength of a severe geomagnetic storm by more than 60%.

The idea has generated enthusiasm as it has rippled through the small world of space weather research. “The thought of changing the interaction never occurred to me,” says Sibeck, who spent decades studying magnetic reconnection. “If I knew that a 100-year disturbance was coming and it would knock out power grids,” he says, “I definitely would want this.”

Craig Rodger, a space physicist at the University of Otago, recalls a senior engineer with New Zealand’s grid operator explaining the idea in a pub during the country’s annual space weather exercise, which have become more important after a 2024 solar storm forced the nation to disconnect some of its transmission lines. “He was excited!” Rodger says.

The system could do more than tame solar storms. The U.S. Department of Defense is exploring ways to clean out radiation that would be trapped by Earth’s magnetic field after a nuclear detonation in space. In February, NRL tested a related technique by releasing barium from a sounding rocket. As the gas ionizes and spreads, it seeds instabilities that can knock high-energy particles out of orbit and into the atmosphere, where their potency dwindles. StormWall could flush out radiation similarly, Walsh says. “It could be one safety net that protects from a few things.”

The engineering challenges may ultimately prove easier than the political ones. A functioning StormWall would depend on accurate space weather forecasts, clear deployment protocols, and international agreement about the deliberate modification of near-Earth space. Walsh had an early conversation with a venture capitalist, who suggested running the system as a subscription service—a nonstarter, given the physics involved, Walsh says. “Either everyone gets it or nobody gets it,” he says.

Of course, like an airbag, StormWall would have to be replaced if deployed. But just as NASA and other space agencies are studying how to protect the planet from asteroids , Walsh says there’s a good argument for fortifying an electronics-dependent society against massive solar eruptions. “If we lose all our power grids and can’t use the internet for 6 years, it would be traumatic.”

In 2020, physicist Ranga Dias published a paper in which he claimed to have discovered a material that acts as a superconductor at room temperature—a longtime dream of materials scientists. Two years later the paper was retracted. In 2023, the University of Rochester (U of R) physicist made an even more striking claim, of a new room-temperature superconductor that works at moderate pressures. That paper was also retracted.

During the uproar, the university conducted three investigations into allegations that Dias had fabricated and falsified data behind those claims and that he had plagiarized parts of a grant application to the National Science Foundation (NSF) to carry out the research. None found evidence of research misconduct.

Then, U of R received a letter from NSF’s Office of Inspector General (OIG) saying it believed the allegations warranted a full investigation. At OIG’s urging, U of R officials assembled yet another panel, which in March 2024 concluded Dias was guilty of fraud and plagiarism. By the fall Dias was no longer on the faculty.

But OIG’s watchdog role over research misconduct has ended, Science has learned. Shortly after President Donald Trump took office in January 2025, it was stripped of its ability to track down scientific miscreants. The change was part of a drastic downsizing of OIG, including the team of Ph.D. scientists responsible for doing those investigations, as Retraction Watch first reported . NSF’s 220-page budget request for 2027, which Trump sent to Congress in April, describes the elimination of 22 OIG positions since Trump took office, a 24% decline.

The resulting “workforce challenges,” according to the budget document, have prevented OIG from investigating research misconduct. Instead, all allegations of research misconduct received by OIG will be referred to the grantee institution to do as it sees fit. If the institution finds the scientist guilty, OIG is supposed to simply send the report up the chain for NSF senior management to adjudicate.

That’s a huge change from how things have worked for 35 years. Created in the aftermath of several academic research scandals in the 1980s, the NSF inspector general’s job is to root out all manner of waste, fraud, and abuse involving NSF dollars. In 1991, NSF explicitly gave OIG responsibility for research misconduct. Since then, it has aggressively pursued allegations that many universities would prefer to sweep under the rug. OIG closely monitored institutions’ investigations and would launch its own, de novo, investigation if it smelled a coverup or thought the university had bungled the job.

The budget request describes the changes as an “interim process” pending a “longer term solution.” An NSF spokesperson declined to answer specific questions about the new policy, but said, “We remain committed to ensuring the integrity of the research we support.”

The former head of the OIG unit that conducted those investigations questions that assertion. “I fail to understand how removing OIG oversight of research misconduct improves the scientific integrity for NSF, its awardees, or its PIs [principal investigators],” says Aaron Manka, a physicist who retired in February 2025 after a 29-year career at OIG. “As I read [the budget], there will be no OIG assessment of allegations, nor will OIG conduct inquiries, or investigations, nor will it assess the institutions’ reports [to NSF].”

Manka left NSF as part of a purge of senior management that included the departure of his boss, Allison Lerner, who had been NSF’s inspector general since 2009. Manka says he oversaw a team of five Ph.D. scientists doing investigations.

Institutions have always had primary responsibility for investigating research misconduct. But NSF’s OIG was known for more aggressive oversight than the Office of Research Integrity within the Department of Health and Human Services, which doesn’t have the same power to subpoena witnesses or documents. “NSF exerts a bit of a firmer hand on the process in terms of what you should and shouldn’t do,” explains Lauran Qualkenbush, head of research integrity at Northwestern University. In the case of alleged plagiarism, Manka notes, his office would often do its own legwork because it has detection software and access to the documents.

That close monitoring of a case also created a sense of urgency, adds Minal Caron, an attorney with Ropes & Gray who works with institutions involved in misconduct probes. “When the OIG is involved,” he says, “there’s a lot more incentive for universities to meet any deadlines NSF has set because of the fear that, if you don’t, NSF could do its own investigation.”

Documents released as part of the Dias case show how OIG could exert that pressure. “Our office reviewed an allegation of research misconduct against one of your faculty members, Ranga Dias, [and] determined there is sufficient substance to warrant an investigation,” OIG investigative scientist Beth Masimore wrote to Stephen Dewhurst, U of R’s interim vice president for research, on 16 March 2023. (The letter was released as part of an unsuccessful suit that Dias filed against U of R claiming its investigation was flawed.) “Please send me by March 31 the policies and procedures you will follow in carrying out this investigation,” Masimore added. “And please note that your investigation should be completed within 180 days.” Her letter also urged U of R to talk to her office about its plans for the investigation and even told staff how to proceed, including not alerting Dias before visiting his office to scoop up lab notebooks and computer files.

Qualkenbush says major research universities such as Northwestern don’t need such handholding. “So no, this will not change how I operate,” she says about taking OIG out of the picture. But she is worried about how long this “interim” policy might remain in effect.

“Is NSF going to just give up oversight [of research misconduct] and leave it entirely to the universities?” she wonders. “The [Trump] administration says it wants to rein in scientific misconduct,” she adds. “But this change seems to be going in the opposite direction.”

The Allen Institute, known for backing large-scale basic biology research such as ambitious “atlases” of cell types in the brain, is expanding its scope to tackle multiple neurodegenerative diseases. The Seattle-based center is already running a study of brains from Alzheimer’s disease patients, but today it announced a plan to scale up its brain-mapping efforts and support the discovery of therapeutic targets across a broad range of disorders.

The 14-year, $400 million project, dubbed the Brain Health accelerator, will use a variety of molecular technologies to analyze postmortem tissue from thousands of brains from both healthy donors and those with diseases including Huntington, Parkinson’s, amyotrophic lateral sclerosis (ALS), and dementia with Lewy bodies. About half its funds come from the institute, with the Bezos Family Foundation, Amazon Web Services, and other partners contributing the rest.

The effort builds on the institute’s ongoing Seattle Alzheimer’s Disease Brain Cell Atlas ( SEA-AD ) study, which has pinpointed cell types and neural circuits damaged even before proteins such as amyloid and tau form the abnormal brain deposits that have long defined the condition’s pathology.

“Most of the field, of course, focuses on the proteinopathy—the amyloid and the tau,” says Allen Institute neuroscientist Ed Lein, who is leading the new effort. Defining, and minutely mapping, affected cell types and circuits makes it possible to start to look for points for intervention, he says, some of which may be relevant across multiple diseases.

Approaches like the Allen Institute’s offer “foundational knowledge,” says John Ngai, director of the U.S. National Institutes of Health’s Brain Research through Advancing Innovative Neurotechnologies Initiative, which supports Allen Institute investigators but is not funding the new initiative directly. “We need to understand all the [brain’s] cell types,” he says: where they are, how they work, and how and why they start to fail.

As part of the new accelerator, the institute is also creating maps of nonhuman primate brains to identify which cell types align between humans and these model species, and determine which are also analogous in mice. More precise matching of human to animal brain circuitry can help lower the risk for drug developers eager to try out drugs or gene therapies aimed at specific cell types, says Lien, who predicts the work of the consortium will lead to an in-human trial of a therapy within 5 years.

The Brain Health accelerator’s budget is quadruple that of SEA-AD. The 29 collaborating institutions, which include universities and disease-focused nonprofits such as EverythingALS, will rely heavily on artificial intelligence to analyze the huge data sets generated by the molecular analysis of donor brain tissue.

Neurologist Bradley Boeve of the Mayo Clinic, who studies frontotemporal dementia (FTD) and is not involved in the new initiative, says it’s “exciting” to see the project moving forward. It “seeks to address many gaps from the cellular and circuit network perspectives” that are pertinent to diseases such as FTD, he says, “with the obvious hope that individualized treatments can advance with even greater urgency.”

The Allen Institute has already shown it can get provocative findings from sophisticated single-cell analyses, says Cristina Sampaio, chief medical officer for CHDI Foundation, a private research organization dedicated to Huntington that is not part of the effort but has been in discussions to join. The biggest challenge for this new project, she says, is “the human factor”—getting a supply of human donor brains alongside detailed clinical and imaging records, and training participating institutions in new ways to preserve them for detailed analysis such as RNA sequencing, which can identify what genes were active in a cell.

Ngai is optimistic that this and other brain-cell mapping initiatives will point the way to novel therapies. “We’re not there yet,” he says, but “we’re reaching an inflection point where researchers have the potential to target just about any cell type in the brain.”

Faced with a large, rapidly expanding Ebola epidemic and no medicines to combat it, health officials in the Democratic Republic of the Congo (DRC) are pondering whether a desperate situation calls for a desperate measure: deploying two Ebola vaccines not designed for the rare virus species causing this outbreak, the Bundibugyo virus (BDBV).

Reports from two expert groups convened by the World Health Organization (WHO) issued on 28 May were lukewarm about the idea, arguing there is scant evidence the vaccines will work. If the shots are used at all, the panels argued, it should be as part of clinical trials aimed at testing their worth. Some experts warn that “mismatched vaccines,” if they turn out to offer little or no protection, could reduce confidence in Ebola vaccines and undermine future vaccination campaigns.

Others, including in the DRC, think the vaccines are the best chance to save lives now. The DRC’s Ministry of Public Health plans to hold meetings with public health officials and researchers this week to decide how to proceed.

The dilemma shows, again, how Ebola has long been neglected, says Glenda Gray, a pediatrician at the University of the Witwatersrand and a member of one of the WHO panels, the Strategic Advisory Group of Experts on Immunization (SAGE). “It’s catastrophic that we don’t have enough data” from immune studies of humans and monkey experiments, says Gray, who adds she’s not speaking for SAGE. “This kind of work should have been done long ago.”

The Ebola outbreak, first detected in the DRC’s Ituri province 2 weeks ago, has already caused 291 confirmed cases and 43 deaths . The virus has spread to two other provinces and neighboring Uganda.

A licensed vaccine against BDBV does not exist. Two separate efforts , one by the nonprofit IAVI and the other by the University of Oxford, are racing to produce one, and today, the Coalition for Epidemic Preparedness Innovations (CEPI) announced it will give an additional $8.6 million to Oxford and $2.3 million to IAVI to speed those projects forward. CEPI also committed up to $50 million to Moderna to manufacture and stage phase 1 trials of a messenger RNA vaccine against BDBV that mirrors the company’s widely used COVID-19 shot.

Separately, Gavi, the Vaccine Alliance today announced it has $40 million for “advanced purchase commitments” to manufacturers, assuring companies will have a market if their products move forward to large studies or even approval.

But it will be months before any BDBV vaccine is ready for testing in humans. The question the two WHO panels— SAGE and a Technical Advisory Group ( TAG )—weighed is whether existing vaccines developed against two other Ebola variants should be used in the meantime.

One of them is Ervebo, an effective and safe vaccine developed by Merck to protect against Ebola Zaire, the variant that has caused 15 of the 16 previous outbreaks in the DRC. Approved by DRC regulators in 2020 , Ervebo contains a weakened, harmless version of the vesicular stomatitis virus (VSV)—which primarily infects livestock—engineered to display the Ebola Zaire surface protein on its outer membrane.

The second vaccine, made by IAVI, uses the VSV system as well but is designed to protect against the Sudan species, which has caused outbreaks in South Sudan and Uganda. That vaccine has yet to prove it’s safe and effective, but IAVI has stocks of it that were prepared for human trials. (The Ebola variants are called “species,” not “strains,” because they are only distantly related. The amino acids that make up these surface proteins differ by up to 40% .)

Encouraging data come from experiments with small numbers of monkeys , which showed that VSV-Zaire (analogous to Ervebo) alone offered some protection against death from Bundibugyo. And using it as a booster shot 14 days after a priming dose of VSV-Sudan fully protected the animals from disease.

The WHO groups also reviewed unpublished lab studies by the French biomedical research agency INSERM, posted as a preprint on medRxiv the day the statements came out. They showed people vaccinated with Ervebo produced antibodies that bound to BDBV, although not as strongly as they bind to Zaire. INSERM immunologist Yves Lévy, who headed the study, says the data are “very encouraging” and suggests the vaccine could offer some protection in the current outbreak.

But TAG and SAGE were not particularly impressed. TAG allowed that Ervebo alone, given as a single shot or as part of a prime-boost combination with the Sudan vaccine, “could be an option,” but concluded that “only weak evidence supports the potential for efficacy.” SAGE says Ervebo should only be used in “controlled research settings.” In a decidedly unenthusiastic conclusion, SAGE said, “Existing evidence neither confirms nor excludes the possibility of some degree of vaccine effectiveness against BDBV.”

Using Ervebo outside of clinical trials, SAGE warns, “could generate a false sense of security among affected communities and responders, potentially … damaging confidence in Ebola vaccination” if people who receive the vaccine get sick.

But Steve Ahuka-Mundeke, a virologist at the DRC’s National Institute of Biomedical Research who will take part in this week’s discussion with the country’s health ministry, says he favors using Ervebo, both in a clinical trial and in “compassionate use,” meaning people could receive the vaccine even though it’s not licensed for BDBV.

People in the DRC are already familiar with the vaccine, Ahuka-Mundeke points out: Three hundred and forty-five thousand received it during an outbreak of the Zaire species between 2018 to 2020 that was also centered in Ituri. And deploying Ervebo in the current outbreak would be a twofer, he notes: “We can protect people from future Zaire outbreaks, and we can assess whether it works or not with Bundibugyo.” He also supports use of the prime-boost strategy, but only in a clinical trial.

In contrast, Armand Sprecher, an epidemiologist and physician with Doctors Without Borders—one of the leading international aid groups helping the DRC respond to the outbreak—says he favors a clinical trial for the prime-boost strategy over Ervebo. “That one has the best evidence base,” Sprecher says. He says the study should focus on health care workers and other front-line workers who are at high risk of infection. His group has found that care providers are especially vulnerable after they take off their protective gear and return home, as sick community members often seek them out for care. Unlike the general population, health workers are easy to contact for the second shot, he adds.

Nancy Sullivan, a virologist at Boston University who previously ran its National Emerging Infectious Diseases Laboratories, published a study 16 years ago showing a prime-boost regimen that delivered Zaire and Sudan surface proteins—using vaccines not in consideration now—solidly protected monkeys from Bundibugyo. She also thinks testing the idea makes sense. Using existing vaccines with “some evidence of cross-reactivity provides the fastest opportunity to save lives now,” Sullivan says.

Trials of the two vaccines would face formidable design problems, says Ira Longini, a biostatistician at the University of Florida and consultant to WHO. “A mismatched vaccine, in my opinion, is very unlikely to have high enough efficacy to even detect it,” Longini says. That means you’d need “extremely high numbers” of subjects to show an effect, he says.

The DRC must make decisions quickly with the limited data that exist, Ahuka-Mundeke says. “We have our heads under the water,” he says. “It’s a difficult time.”

A federal judge today temporarily blocked the U.S. National Science Foundation (NSF) from transferring control of a supercomputing facility operated by the National Center for Atmospheric Research (NCAR), the world-leading climate research lab, to a different institution. Transferring the facility risked disrupting the center’s infrastructure and could potentially degrade its climate forecasting and scientific capabilities, ruled Colorado District Court Judge R. Brooke Jackson.

The ruling marks the first tangible victory in a lawsuit brought against NSF by the universities that collectively manage NCAR. Late last year, NSF, at the direction of Russell Vought, director of the Office of Management and Budget (OMB), announced it would “break up” NCAR, which Vought targeted for its role in “climate alarmism.” It is also widely believed NCAR was targeted as part of the White House’s now-successful campaign to pressure Colorado to release from prison Tina Peters, the former county clerk who was convicted of breaching election security systems to search for fraud in the 2020 presidential election.

NSF earlier this year announced its intent to transfer control of the NCAR-Wyoming Supercomputing Center, an important resource for academic researchers running weather and climate models, potentially to the University of Wyoming. NSF has also threatened to transfer NCAR’s research aircraft and to break up its academic centerpiece at the Mesa Lab in Boulder, Colorado. Seeking to block these changes, in March the University Corporation for Atmospheric Research (UCAR), which manages NCAR, sued the government , seeking an injunction on NSF’s actions.

Shortly after filing the suit, UCAR sought a preliminary injunction that would specifically protect the supercomputing center. In its defense, the government contended it had not actually made a final decision to transfer stewardship of the facility, even though an NSF program manager in March told NCAR’s leadership the effort to move the supercomputing center to the University of Wyoming was “a decision made.”

Jackson found that the government’s decision to transfer the supercomputing center was arbitrary and capricious, given that NSF “offered no explanation for its decision” and failed to abide by its own processes for soliciting public feedback before making the decision. He also found convincing UCAR’s argument that NSF’s actions had caused harm by causing several senior staff to leave the supercomputing center. NSF, meanwhile, provided no concrete evidence that keeping the supercomputer under NCAR’s management would cause harm. “Indeed,” Jackson wrote, “it would be difficult to do so on this record given that the agency has identified no deficiency in UCAR’s performance as steward of the [center].”

The ruling is only the first skirmish in what is likely to be a much longer legal battle. UCAR is still seeking a broad injunction against NSF making additional changes at NCAR, and the state of Colorado has also sought to prevent any changes in a separate case. It also remains to be seen whether the release of Peters by Colorado’s Democratic governor, Jared Polis, will change the White House’s strategy.

Not long after President Donald Trump started his second term, many scientists funded by the U.S. National Institutes of Health (NIH) began to be asked by the agency to strip certain words from their grants, sparking cries of research censorship and politicization. Now, their projects are facing an even higher level of scrutiny from the Trump administration. According to documents viewed by Science , all grants approved by NIH for funding are now going through an extra screening at its parent body, the Department of Health and Human Services (HHS). Staffers there who may be political appointees and not necessarily subject matter experts sometimes ask for substantive changes in the research.

Some NIH career employees involved in the grant-review process say the HHS requests are unprecedented and alarming. “The U.S. scientific community and the broader American public should be deeply concerned by the fact that HHS is overriding peer review to require changes to research scope, design, and language,” says Jenna Norton, an NIH program officer who stresses she is speaking in her personal capacity, not on behalf of NIH. (Norton was put on leave last fall after leading protests against Trump administration actions, but has since been reinstated.)

The extra layer of review includes both new proposals that NIH deemed worth funding and ongoing grants awaiting their annual payment. HHS reviewers in several cases have even asked for changes in the work to be conducted near the end of a multiyear grant when a course correction would be impossible, according to NIH staffers who spoke with Science on the condition of anonymity to protect their jobs.

The number of NIH grants in which HHS has demanded changes is unclear, although the practice does not appear widespread. And Science has not learned of any specific proposal that was not funded as a result. Still, the extra HHS review adds to other factors delaying NIH’s grant decisions, including staff shortages and a backlog from last year’s federal shutdown. The agency is far behind its usual pace of grantmaking at this point in the fiscal year—although NIH says it has picked up the pace and will spend all of its budget before this fiscal year ends on 30 September.

An HHS spokesperson said on background that NIH staff are still making final grant decisions, and that requests like those coming from HHS are routine. “It is not unusual for NIH staff to suggest changes to research plans in the interest of strengthening new or ongoing research programs,” the spokesperson said.

Even before this new HHS review was instituted, NIH grant proposals and renewals had begun to receive extra scrutiny , starting around when health economist Jay Bhattacharya took over as director in April 2025. After a proposal is peer reviewed and receives institute approval for funding, or an ongoing grant is cleared for its yearly payment, an NIH office now uses a computational tool to scan publicly available language about each grant for politically sensitive words such as “gender,” “equity,” and “diversity.” It also looks for those words in the nonpublic description of the research aims. (Certain multiyear grants seeking their annual renewals are currently exempt from the review.)

That office then sends spreadsheets of flagged grants to each institute, which must then renegotiate the grants with the applicants to replace or remove the words. Some proposals flagged by the computational tool are later deemed false positives because a term such as “diversity” was used to describe something noncontroversial, such as a cell population.

It now appears that since mid-April, grants have been going to HHS for another layer of review. One internal NIH communication Science has seen states that awards approved for funding are sent at the end of each day to “HHS Counselor,” apparently a reference to the office of HHS Chief Counselor Chris Klomp. NIH staffers say HHS, too, seems to be using a computational tool to screen the grants. But in addition to flagging sensitive terms, HHS is in some cases asking for substantive changes to the research.

For example, according to documentation seen by Science :

• In a project studying factors contributing to depression using many years of observational data from large cohorts, the HHS review recommended adding genetic influences. The project was in its fourth year, when such changes were likely not feasible.
• HHS asked that an ongoing grant proposing to study obesity in a minority group more clearly explain why certain “stressors” might influence weight. The grant was approved after the investigator added more stressors and other potential factors and described other health outcomes in addition to obesity.
• An ongoing study of a health matter in a certain occupation focused on a minority population. The commenter asked whether a specific health problem was more common with this group or its members responded differently to an intervention. The investigator added explanations in response.

In another case described to Science by an NIH program officer, HHS requested the investigator add a new analysis to a study examining factors contributing to obesity in a minority population.

Sometimes, the demands make no sense, NIH staffers say. For example, HHS wanted a scientist training grant in its final year to add a clinical trial—work these awards by definition do not support. “It’s an absolutely bonkers comment,” a program officer says.

When the World Health Organization (WHO) announced on Thursday its recommendations for treatments and preventive measures to test in the current Ebola outbreak, a lot of the candidates were familiar: vaccines in various stages of development (though none ready to test are specific to the rare Ebola virus variant known as Bundibugyo that’s currently spreading), and antivirals and monoclonal antibodies for those who fall sick. But one thing is new: a trial for a 10-day course of pills intended to protect people after they have been exposed to Ebola.

Researchers and health officials hope the approach can slow the outbreak, which has now caused more than 1000 suspected cases including more than 200 deaths in the Democratic Republic of the Congo (DRC) and Uganda. “Since we don’t have vaccines, this is something you can do immediately that could be effective,” says Ira Longini, a biostatistician at the University of Florida.

The preventive strategy, known as postexposure prophylaxis (PEP), mirrors what has proved successful for other infectious diseases. Here, researchers will test an experimental antiviral called obeldesivir. It would be the first controlled trial to evaluate antivirals as a preventive measure during an Ebola outbreak.

At the moment, direct contacts of Ebola patients get visited daily to check for symptoms, but that’s it, says Armand Sprecher of Doctors Without Borders. “All you’re doing every day is saying, ‘Are you sick yet? Are you sick yet?’” Sprecher says. If obeldesivir works, it would not only prevent cases and thus spread, but might also facilitate contact tracing by motivating more contacts to come forward, he says. “That could be a game changer.”

Gilead began to develop obeldesivir early in the COVID-19 pandemic as an oral counterpart to its antiviral remdesivir, an intravenous drug that was widely used against the novel coronavirus SARS-CoV-2. Clinical trial results for remdesivir against Ebola viruses have been mixed. It did not clearly reduce mortality among people with the Zaire strain of Ebola in the PALM trial, conducted during a devastating outbreak in the DRC that started in 2018. Still, some evidence suggests it can help Ebola patients if given early enough after infection.

The idea for the current prevention trial was also born from that same outbreak. In 2019, Marie Jaspard, an infectious disease specialist at Sorbonne University, was working in Beni in the eastern DRC when a health care worker asked how to help a woman who had been caring for her ill husband at home and then learned he was infected with Ebola. A vaccine for Ebola Zaire was available, but it would take several days for its protection to kick in, too late for the woman. “The research question came from this exact moment where I had to answer this colleague,” Jaspard says. “To me there was clearly a gap in the care.”

So Jaspard, with Congolese virologist Placide Mbala and other researchers, started an emergency-use program offering one of two experimental antibody drugs against Ebolaviruses, called mab114 and REGN-E3B, to people who had had direct contact with symptomatic patients. Both had just been shown in the PALM trial to reduce mortality in patients infected with Ebola Zaire, but were not yet approved as therapies. They might be even more powerful virus fighters if delivered before symptoms emerge, the researchers reasoned.

None of the 23 people who received one of the two drugs over the next 6 months got sick. But it is unclear how many infections would have been expected in the group, so after the outbreak ended, Jaspard put together a consortium, including the National Institute for Biomedical Research in Kinshasa, DRC, and the medical humanitarian organization ALIMA, that designed a larger and more rigorous test—a trial called EBO-PEP—to be ready to use in the next Ebola Zaire outbreak.

More recently, Jaspard and her colleagues updated the EBO-PEP protocol to include another candidate: obeldesivir, which has the benefit of being taken as a pill. Like remdesivir, it is transformed in the body through a series of steps into the active molecule GS-443902, which mimics the structure of ATP, a molecule viruses need in order to replicate. “It’s like somebody sticking a piece of a cardboard into the photocopier, and it jams things up,” Sprecher says. But unlike remdesivir, obeldesivir can be absorbed in the gut before it’s turned into the active molecule.

Tests of obeldesivir as PEP in monkeys have been encouraging. Between 80% and 100% of animals survived when they received the treatment 24 hours after being injected with the Zaire and Sudan strains of Ebola, or with the related Marburg virus. Without the drug, the animals didn’t stand a chance, says virologist Thomas Geisbert of the University of Texas Medical Branch, who led the studies. “They all die, and they die quickly.” But, he adds, “Guess what, we didn’t test [it against] Bundibugyo.”

Jaspard and her colleagues are now adapting the EBO-PEP protocol for the current outbreak. They face some thorny issues, including what the trial should offer people in the control group. The original trial protocol, designed for an Ebola Zaire outbreak, gives all participants the Zaire vaccine, randomizing them to vaccine-only or vaccine-plus-antiviral groups. Some researchers think even with Bundibugyo, the Zaire vaccine is likely to have a short-term protective effect because the vesicular stomatitis virus used to deliver Ebola’s surface protein generally ramps up the immune system. But in the current outbreak, WHO has discouraged trialing a vaccine that is designed to protect against a different pathogen.

At a meeting of the trial consortium on Tuesday, Sprecher says Vasee Moorthy from WHO’s Office of the Chief Scientist “made it clear that it was not up for discussion.” The agency hasn’t publicly shared its reasoning, but in an email Moorthy noted the limited data on the effectiveness of existing vaccines against Bundibugyo.

“If we had data to support vaccination (and maybe we will get it at some point) then we would be doing something for everyone. That would be my preference,” Sprecher says. But he also points out that most contacts of an infected person will never develop Ebola, meaning many of those receiving the drug would only experience the side effects. It will be up to health authorities in the DRC and Uganda to decide whether they accept a trial design where the control group only gets placebo pills.

Another question is how best to distribute obeldesivir. Because family members of an infected person might live in the same house but be randomized to different arms of the study, researchers want to make sure they don’t share or mix up their pills. With doses required twice a day for 10 days, it’s not possible for health workers to observe participants taking each pill. “The plan is to have a community engagement team that will go to the houses of participants every day and give them the drugs,” Jaspard says. “We will see if it’s feasible.” In Uganda this is not a problem, because contacts are usually separated and taken to an isolation center for 21 days, says Pauline Byakika-Kibwika, a researcher at Mbarara University of Science and Technology who along with Jaspard and Mbala is a principal investigator of the study.

The study is meant to only include people at the highest risk of developing Ebola, but it’s not easy to define that group. The team has decided to focus on those who had contact with confirmed Ebola patients with “wet” symptoms such as bleeding or vomiting, those who had contact with the body of an Ebola victim, and those who got a needle-stick injury while treating an infected patient. “I guess that it’s simple enough at this stage,” Jaspard says, “and we worked a lot with the countries to make sure this is acceptable.” Experience from previous work in the DRC suggests there is usually a ring of roughly 20 people who have had direct contact with a patient, Longini says. He has made some calculations about how many participants would be needed to prove whether the preventive strategy works in the trial. If obeldesivir has a very strong protective effect, that could be clear from just a few dozen rings in each of the two arms, he says.

Nearly 7 years after her first attempt to test Ebola prevention drugs, Jaspard is stunned by the sudden interest in the trial. “I’ve kept this project for all these years, despite the fact that a lot of people were saying it’s not useful, you don’t need to do this, go away,” she says. Longini admits he did not think a PEP study was crucial in Ebola until now. “It was not something I thought we’d have to resort to,” he says. “I thought we would have the vaccines ready when these things happen.”

The White House today announced it plans to make major changes in how the U.S. government manages research funding. Some scientists are horrified, saying the proposal from the Office of Management and Budget (OMB) would replace peer review with the judgment of nonexpert political appointees sworn to carry out the wishes of President Donald Trump. Others see a silver lining in the 412-page document because it preserves, for now, how universities are reimbursed for indirect costs, the billions of dollars they spend each year to support federally funded research on their campuses.

It’s too early to know how grantmaking will ultimately be changed by the proposed rule, which codifies many Trump executive orders relating to federally funded research. The document, published today in the Federal Register , not only declares that “senior appointees” at federal agencies must sign off on awards, but would ban many foreign collaborations and cut federal support for publishing costs. OMB has given the community 45 days to comment on the notice before it issues a final rule this fall. But that uncertainty hasn’t stopped those who follow U.S. science policy from weighing in on its significance.

“What OMB is proposing is not a reform of grants management,” Elizabeth Ginexi, a former program officer at the National Institutes of Health (NIH), writes in a Substack post . “It is a vehicle for complete political control of science … over every stage of the federal science funding lifecycle.” Representative Zoe Lofgren (D—CA), a leading critic of the Trump administration’s research policies, calls the proposal “a dystopian move that would destroy what remains of merit-based review.”

At the same time, professional and higher education organizations that have lobbied OMB not to overturn the current system of indirect cost reimbursement are grateful for its decision to maintain the status quo. “It provides us some breathing room for further discussions about a better solution that is fair and that increases accountability and transparency,” says Debbie Altenburg of the Association of Public and Land-grant Universities.

Today’s announcement suggests the idea of capping reimbursement rates at a much lower level, as several federal agencies tried unsuccessfully to do last year, is still alive. “OMB may consider issuing a request for information on this topic in the future,” it says.

OMB has also proposed essentially preventing scientists from tapping their grants to defray the cost of publishing their papers in open-access form.

These article-processing charges (APC), which average more than $3000 per paper, are often paid to satisfy federal requirements for public access to findings generated by government grants. But OMB argues that such fees “may serve institutional, professional, or reputational interests rather than the specific objectives of the federal program” and thus, should generally not be allowable expenses. If adopted, the new rule could shift the burden of paying APCs from the federal government to institutions or individual researchers. Some scientists whose research is not supported by federal funding are reportedly paying APCs out of their own pockets.

The proposed regulation would also give additional legal weight to several Trump executive orders last year that banned research to foster diversity, equity, and inclusion in science on the grounds that it favored certain groups and slighted others. It says tighter oversight by political appointees is needed to prevent research agencies from “promot[ing] a woke policy agenda that did not reflect the values of the vast majority of the American public.”

Trump has directed federal agencies to ban such research and ensure that all funded projects are “aligned with the administration’s priorities.” Ginexi and other critics say that approach distorts a process that is supposed to reward the best scientific ideas.

At NIH, leaders have maintained that an August 2025 executive order cited in today’s proposal requiring political appointees to sign off on grants does not mark a change from past practice. Asked last fall about the order during a meeting of NIH’s top cancer advisory board, NIH Director Jay Bhattacharya said, “The NIH director has always been a political appointee, and the NIH director sets priorities for the entire of the NIH."

Neal Lane, who led the National Science Foundation (NSF) under former President Bill Clinton before serving as his science adviser, believes there’s an important distinction between funding good science to further a presidential initiative, which is what Bhattacharya says NIH is doing, and rejecting equally good research ideas because they don’t square with a president’s political views.

“NSF makes grant decisions based on the deep knowledge of experts in the field, not on whether it meets an ideological agenda,” Lane says. “That’s what merit review is all about. Replacing it with top-down decision-making will destroy that process and result in bad science being funded.”

Altenburg says she can’t predict whether OMB will modify the proposal based on the comments it receives. But she is encouraged by the administration’s willingness to follow the law by soliciting public comments. “They are required to respond to every comment, so we’ll see what happens,” she says.