Trump Administration Works to Rebuild AI Bioweapons Defenses After Drastic Staffing Cuts

Trump aims to rebuild defenses against bioweapons as AI fears rise

(Illustration by Chelsea Conrad/For The Washington Post; iStock) The White House is working to strengthen safeguards against biological attacks after staffing cuts early in the second Trump administration depleted expertise in a field where officials increasingly fear artificial intelligence could make it easier to create deadly pathogens. By the end of President Joe Biden's term, the White House had built a team of as many as 30 people focused on biological security, according to three former government officials who worked on the issue -- two for Biden and Trump, and one just for Biden -- and spoke on the condition of anonymity to avoid professional retaliation. This specialized staff was charged with coordinating the web of federal agencies that deal with biohazards and provided scientific guidance to top officials. That team rapidly dwindled after Trump returned to office amid cuts and reorganizations, said the three former officials, as well as a fourth former official, who also worked for both administrations. At times, nobody at the White House was dedicated solely to biosecurity, according to three of the former officials. The losses have come as the administration has delayed replacing key Biden-era directives. It revoked a 2023 executive order on AI calling for stronger biological safeguards and cast doubt on a policy issued the following year designed to require companies to screen buyers seeking genetic components that could be used to create biological agents -- widely seen by experts as a basic pillar of the nation's defenses. Trump directed officials to produce a revised version by August 2025. More than a year later, it has yet to appear, even as AI companies say their most advanced models are gaining increasingly sophisticated biological capabilities alongside the hacking skills that have alarmed officials this year. "They have kneecapped themselves in their ability to create any policy that would have staying power," said one of the former experts, who recently worked directly on outbreak response policy. "That's the risk, they're not playing the long game. The whole purpose of having the technocrats there is to advise political leaders on the nuanced tradeoffs of taking one path or another." In the absence of a large team, the White House has relied on a small revolving cast of employees. They have included a junior fellow without a security clearance; a part-timer from the research organization Rand; a pair of CIA staffers with little expertise in the field; and an Air Force doctor who had been a consultant to the White House medical office, according to the three experts. A White House official said the administration was working to strengthen the Biden-era policy, calling it inadequate and technically limited despite the staff devoted to developing it and did not address questions about the staffing changes. "The Administration's commitment to addressing AI and biological safety risks is clear and ongoing," the official said in an email. "The White House has been developing policy for over a year, including in our AI Action Plan last summer." The Trump administration has acknowledged the biological security risks AI poses while touting the potential for the technology to help keep the nation safe. Trump told foreign leaders gathered at the United Nations last year that he wanted to prevent biological weapons disasters and said AI was a tool to help. At the Pentagon, a former health official who led Operation Warp Speed -- the effort during the first Trump administration to rapidly produce a vaccine for the coronavirus -- is involved in efforts to use AI to speed the development of new defenses against future pandemics and to deter biowarfare. Part of the work involves analyzing billions of tweaks to antibodies to assess how to respond to new forms of the coronavirus. But that work is taking place in a government that analysts and the former experts say has been drained of expertise amid broader cuts to federal public health and science funding. The White House removed all the staff from the Office of Pandemic Preparedness and Response Policy, an organization created to lead the nation's response to the kind of major outbreak that the most dangerous kind of bioweapon could trigger. The office was resurrected as Ebola spread in Africa this spring. A National Security Council unit dedicated to health security was also shut, two of the former experts said, and parts of the Office of the Director of National Intelligence and Department of Homeland Security responsible for biological threats were broken up as part of broader reorganizations. A scientist who worked on biotechnology and national security issues in both the first Trump and Biden administrations said the federal government is not taking a leading role in coordinating a response to the risks posed by the rapidly evolving technology. "Every time an administration changes over there's a lot of continuity that just gets disrupted," said the scientist, who spoke on the condition of anonymity because they were not authorized by their employer to speak publicly. In the transition from Biden to Trump, they said, "when it comes to biosecurity, that happened in a very disruptive way." There has never been one federal agency responsible for biosecurity, a sprawling field that encompasses efforts to secure research labs, aspects of public health work and international diplomacy, and which is increasingly focused on the role of AI. But an analysis of Trump's most recent budget request by the Council on Strategic Risks, a Washington-based security-policy think tank, concluded that the administration's rhetoric on biosecurity was not matched by its funding proposals. In a separate analysis, the group found that spending across multiple government departments peaked in 2024, before declining slightly in Trump's first year -- with the administration proposing further cuts. "General programmatic cuts and agency-specific reorganizations have restricted and eliminated key programs, clouding budgetary transparency and weakening US deterrence efforts," the council wrote in a blog post. When Trump returned to office last year, his team was focused on eliminating barriers to the development of AI technology. That approach began to shift this spring when a new generation of models proved to be capable of hacking into computer networks, leading administration officials to embrace the kind of government oversight of yet-to-be released tools that they had once spurned. There are now signs that officials are taking additional interest in the potential dangers. That has included a recent string of briefings on biological risks between OpenAI and the White House and half a dozen government agencies, according to a person familiar with the meetings, who spoke on the condition of anonymity to describe their private nature. Politico reported this month that the White House was launching a new initiative involving its cybersecurity office and a top official at the Department of Health and Human Services. Leaders in the AI industry have long said that their tools being used to enable a biological attack ranks among their top fears. "I'm getting very frustrated," said David Relman, a professor of medicine at Stanford University who worked on pandemic response in the Biden White House. The newest AI models, such as Anthropic's Mythos and OpenAI's GPT 5.6, have continued to get better at high-level biological research. But just as the models' ability to write computer code has made them good at hacking, their biological capabilities have stoked fears that they could be used to help build a new bioweapon, set off a new pandemic or aid even a relative amateur in carrying out a biological attack. This month, the journal Science published the results of a successful effort to use AI to engineer new viruses. They were designed to infect bacteria rather than humans, but the outcome was a clear sign of what is becoming possible. As early as last summer, researchers at OpenAI determined that they were close to reaching "high" levels of capability in biology, which the company defines as capabilities that "significantly increase existing risk vectors for severe harm." The company briefed the Trump administration and hosted more than 50 government experts for a summit at its San Francisco headquarters that July to discuss how its systems could be used for biological defense. (The Washington Post has a content partnership with OpenAI.) The company said it has continued to work with the administration, aiming to provide officials with access to a special version of GPT for biological defense that it calls Rosalind. "The Administration has increasingly prioritized this important issue as AI capabilities have expanded, and we welcome their partnership," an OpenAI spokeswoman said in an email. Despite the growing concern, scientists remain divided on how serious the risks are, how soon they might materialize and how the nation should prepare. Large language models that power chatbots could be helping terrorists to do tricky lab work, and specialized biological design tools could aid in the creation of new kinds of deadly pathogens not readily recognized by current methods of surveillance. In October, researchers at Microsoft detailed how they had patched gaps in detection systems after using AI to engineer digital versions of thousands of toxins that were not caught by the existing safety net. A report by Rand concluded that attackers have a significant advantage over defenders when it comes to biological incidents. Someone carrying out an attack could make multiple efforts, while authorities need to be able to tackle every threat because even one "could result in catastrophic consequences." Leading American AI firms have sought to lock down their models, blocking them from answering questions about risky biological research and providing more capable systems only to trusted partners. But AI guardrails are imperfect and OpenAI said in July that its testing of leading Chinese models, which are made freely available, indicates that they do not have the same kinds of protections built in. "I think we have a massive collective-action problem in front of us," said Relman, the Stanford professor. AI systems have gone on runaway hacking sprees in recent months but, for now, experts say the barrier between the digital world and the biological one presents a major hurdle -- and an important line of defense. Yet that could begin to change as companies seek to automate laboratories. "The ultimate fears are a ways out," said Crystal Grant, a researcher at the Council on Strategic Risks. "We have a lead time before they're here to prepare ourselves." Because the same abilities can allow AI to be used for dangerous and helpful biological research, many scientists also fear their work becoming too tightly regulated. The Trump administration recently released a new policy on what it calls dangerous gain of function research. It's a long-standing approach involving tinkering with pathogens in order to better understand how to defend against them. But it has become increasingly controversial in the wake of the coronavirus pandemic. The Trump administration has embraced the theory that the virus escaped from a Chinese lab that was undergoing gain of function work. Some scientists say the new policy is vague and could impede legitimate research. In the private sector, researchers complained that biological guardrails Anthropic put on its most powerful model were too strict. The company loosened the restrictions this month. Paul Friedrichs, a retired Air Force general, said the need to get the balance of safety and progress right underscored the need for strong leadership and expertise in the federal government. Friedrichs, who ended his government career leading the White House pandemic office under Biden, said the cuts in the Trump administration have "unquestionably" left the nation less well prepared and at risk of squandering an incredible opportunity. "We're on the cusp of curing some of the banes of humanity," Friedrichs said.

Trump administration works to strengthen defenses against potential AI bioattacks - report

The Trump administration is working to strengthen security against biological attacks after staffing cuts targeted experts in the field early in US President Donald Trump's second term, as artificial intelligence raises new fears about synthetic pathogens, The Washington Post reported on Monday. By the end of former president Joe Biden's term, some 30 people were on a team focused on biological security, WaPo wrote, citing three experts who served as government officials. The team coordinated the federal agencies dealing with biohazards and provided scientific guidance to top officials. When Trump returned to office, the team was cut, and at times, nobody at the White House was dedicated solely to biosecurity, according to the report. The administration also revoked a 2023 executive order on AI calling for stronger biological safeguards and cast doubt on a 2024 policy that required companies selling genetic components that could be used to create biological agents to screen their customers, a policy multiple experts have called for. While Trump called for a revised version by August 2025, no new order has appeared as AI companies say their models are gaining increasingly sophisticated biological capabilities, WaPo wrote. Administration touts benefits of AI The administration has also focused on the potential benefits of AI in these fields. Trump said that AI could be a tool to aid in preventing disasters caused by biological weapons while speaking to foreign leaders at the United Nations last year and WaPo reported that a former health official who led the effort to rapidly produce a vaccine for COVID-19 during the first Trump administration is now at the Pentagon, working on ways to use AI to develop new defenses against potential pandemics and to deter biowarfare. However, the departments dealing with biosecurity have been severely depleted. In addition to the biosecurity team at the White House, the administration also shut down the Office of Pandemic Preparedness and Response Policy and a National Security Council unit dedicated to health security. Parts of the Office of the Director of National Intelligence and the Department of Homeland Security responsible for biological threats were also broken up in the administration's reorganization, the experts said. The Office of Pandemic Preparedness and Response Policy was reopened in the spring following the start of the Ebola outbreak in the Democratic Republic of the Congo, but experts still believe that is not enough, according to WaPo. A White House official told WaPo that the administration was working to strengthen Biden-era policy around biosecurity, calling it limited and saying it did not address questions regarding staffing changes, WaPo wrote. "The Administration's commitment to addressing AI and biological safety risks is clear and ongoing. The White House has been developing policy for over a year, including in our AI Action Plan last summer," WaPo cited the official as saying in an email. OpenAI briefs White House on biological risks Recently, OpenAI and the White House and half a dozen government agencies have held multiple briefings on biological risks, WaPo cited a person familiar with the meetings as saying. Earlier in the month, Politico reported that the White House was launching a new initiative involving its cybersecurity office and a top official at the Department of Health and Human Services. OpenAI also hosted over 50 government experts at a summit in July 2025 to discuss its system's potential use for biodefense, and has since said it has continued to work with the administration to provide a special version of GPT called Rosalind, designed for biodefense, WaPo wrote. "The Administration has increasingly prioritized this important issue as AI capabilities have expanded, and we welcome their partnership," an OpenAI spokeswoman told WaPo in an email. While some scientists call for greater regulation, others fear it may impede important research. Leading US AI companies restrict their models, but Anthropic loosened its guardrails on its most powerful model this month after repeated complaints from researchers.