U.S. AI Safety Institute

The U.S. AI Safety Institute (AISI) is a body within the National Institute of Standards and Technology (NIST) of the Department of Commerce, created in November 2023 at the direction of President Joe Biden to carry out the tasks his executive order on artificial intelligence assigned to Commerce.¹² In June 2025 the Donald Trump administration renamed it the Center for AI Standards and Innovation (CAISI).² NIST describes CAISI as "industry's primary point of contact within the U.S. government to facilitate testing and collaborative research related to harnessing and securing the potential of commercial AI systems."³

Leadership Under Raimondo

On February 7, 2024, Commerce Secretary Gina Raimondo named Elizabeth Kelly, then special assistant to the president for economic policy, as the institute's inaugural director and Elham Tabassi, NIST's chief AI advisor, as chief technology officer. Kelly said that "our first priority will be executing the tasks assigned to NIST in President Biden's executive order," and that she looked forward "to building the institute as a long-term asset for the country and the world."¹

On March 7, 2024, VentureBeat reported, citing unnamed sources, that NIST staff members were threatening to resign over the expected appointment of Paul Christiano, whom it described as an "effective altruist," that the hire "was allegedly rushed through the hiring process without anyone knowing until today," and that it had come directly from Raimondo. The staff concern was that his ties to effective altruism and longtermism "could threaten the institute's objectivity." Divyansh Kaushik of the Federation of American Scientists defended him as qualified for the executive order's work on chemical, biological, radiological and nuclear risks.⁴

On April 16, 2024, Raimondo named Christiano head of AI safety, Adam Russell chief vision officer, Mara Campbell acting chief operating officer and chief of staff, Rob Reich senior advisor, and Mark Latonero head of international engagement, saying: "we need the top talent our nation has to offer."⁵ The Commerce Department's release stated that Christiano would "design and conduct tests of frontier AI models, focusing on model evaluations for capabilities of national security concern," and noted that he had "launched a leading initiative to conduct third-party evaluations of frontier models, now housed at Model Evaluation and Threat Research (METR)." Russell came from the AI division of the University of Southern California's Information Sciences Institute and had been a program manager at the Defense Advanced Research Projects Agency and at the Intelligence Advanced Research Projects Activity and had helped establish ARPA-H. Campbell had been deputy chief operating officer of the Commerce Department's Economic Development Administration. Latonero had been deputy director of the National AI Initiative Office at the White House Office of Science and Technology Policy. Reich, a professor of political science at Stanford University and associate director of its Institute for Human-Centered AI, took public-service leave to advise the institute and "lead engagement with civil society organizations."⁶

Christiano had founded the Alignment Research Center, to which Open Philanthropy gave 1,515,000 dollars in 2022, and was one of the original trustees of Anthropic's Long-Term Benefit Trust, which he left that month to take the post.⁷⁸ Reich had been a director of the Clear Fund, the legal entity of GiveWell, on its returns for 2013 through April 2019, on the same board as Holden Karnofsky and Cari Tuna; GiveWell's own board listing of 2018 identified him as "a Professor of Political Science at Stanford University."⁹¹⁰ The Action Fund of the Center for AI Safety stated that it had led a bipartisan congressional effort that "secured $10 million to establish" the institute.¹¹

Renaming

The Daily Signal reported in April 2026 that Commerce Secretary Howard Lutnick had said at the time of the June 2025 renaming: "For far too long, censorship and regulations have been used under the guise of national security. Innovators will no longer be limited by these standards," and that "CAISI will evaluate and enhance U.S. innovation of these rapidly developing commercial AI systems while ensuring they remain secure to our national security standards."² NIST's page for the center lists among its tasks unclassified evaluations of AI capabilities posing national-security risks, in which "CAISI will focus on demonstrable risks, such as cybersecurity, biosecurity, and chemical weapons," assessments of "U.S. and adversary AI systems," including "the possibility of backdoors and other covert, malicious behavior," coordination with the Department of Defense, the Department of Energy, the Department of Homeland Security, the Office of Science and Technology Policy and "the Intelligence Community," and representing U.S. interests abroad "to guard against burdensome and unnecessary regulation of American technologies by foreign governments." Its published work in 2026 included evaluations of DeepSeek V4 Pro and of two models from the Chinese company Z.ai, and a joint assessment with the British AI Security Institute of Moonshot AI's Kimi K3.³ In September 2026 the OpenAI Foundation described Christiano as a senior technical advisor at CAISI when it appointed him to its board.⁷

Leadership Turnover, 2026

Collin Burns, the first-listed author of OpenAI's 2023 "Weak-to-strong generalization" research, on which OpenAI's president Greg Brockman was listed as a contributor, was selected in April 2026 to direct CAISI.²¹²¹³ Accounts of his departure differ. The Daily Signal, citing sources familiar with the matter, reported on April 23, 2026 that Burns, "a former researcher at Anthropic and OpenAI, was tapped to lead" the center, "but while Burns was in the onboarding process, the Commerce Department selected a different person to lead the organization," Chris Fall, who had directed the Department of Energy's Office of Science in the first Trump administration. It noted that Anthropic had declined the Pentagon's request for unrestricted use of its technology and that Secretary of War Pete Hegseth had declared the company a "supply chain risk to national security."² A letter of August 13, 2026 from Representative Ro Khanna to Lutnick, published by the Democratic side of the House Select Committee on the CCP, stated instead that Burns "was ousted just four days into his tenure as CAISI's leader," adding: "Many believe this was driven by the administration's hostility toward his former employer."¹³

On July 20, 2026, CNBC confirmed that Fall had resigned about three months after being picked; a Commerce spokesperson, Kristen Eichamer, said that NIST's director, Arvind Raman, would serve as acting director and declined to say why Fall had left. The same report noted that David Sacks had stepped down as White House AI and crypto czar in March 2026 without being replaced.¹⁴ Khanna's letter stated that the Trump administration's AI Action Plan had directed CAISI to evaluate Chinese frontier models "for alignment with Chinese Communist Party talking points and censorship," and that CAISI had reportedly stopped publishing findings on content manipulation in its latest evaluations of Chinese models.¹³

In September 2026 the publication Effort, founded by Brian Chau, wrote, citing a former employee of the Foundation for American Innovation, that Sophia Brown, formerly strategy and partnerships director of the Horizon Institute for Public Service, had been hired into CAISI under the Intergovernmental Personnel Act through that foundation.¹⁵