Chris Fall spent three months running the federal government’s newest AI oversight agency. On Monday, he resigned, and neither he nor the Commerce Department offered a public reason why.
The Commerce Department confirmed the departure to multiple outlets, with spokesperson Kristen Eichamer saying only: “Following Chris’s departure, NIST Director Dr. Arvind Raman will continue to oversee CAISI and will serve as Acting CAISI Director.”
🚨 BIG BREAKING
Chris Fall just resigned as head of the US Center for AI Standards and Innovation… after ONLY 3 months.
This is the office that secretly stress-tests unreleased frontier models from OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI for major security… pic.twitter.com/zSrc5jNInZ
— XP SIGNAL (@XPSIGNAL) July 20, 2026
Raman, the former dean of engineering at Purdue University, had been sworn in to lead NIST just three weeks earlier, on June 30. Reuters reported that a White House spokesperson declined to comment on Fall’s exit entirely.
Stay up-to-date with the latest news!
Subscribe and start recieving our daily emails.
Even getting Fall into the job took an extra step.
Commerce’s first choice had actually been Collin Burns, who’d previously worked as a researcher at both Anthropic and OpenAI. Burns never made it past onboarding, though, and Fall, a first-term Trump administration veteran who once ran the Department of Energy’s Office of Science, ended up with the job instead. All of that happened in late April, meaning CAISI burned through its first intended director before its actual director had even started, and now the actual director is gone too.
What CAISI is supposed to do is genuinely significant.
The agency, rebranded from the U.S. AI Safety Institute, is responsible for building federal testing frameworks for frontier AI models, evaluating national security risks in advanced language systems, and setting industrial safety standards for how AI gets deployed across the country.
That mandate has already produced friction this year.
In May, CAISI announced new pre-deployment testing agreements with Google DeepMind, Microsoft and xAI, extending arrangements the agency had already established with OpenAI and Anthropic back in 2024. The official announcement read: “Through these expanded industry collaborations, CAISI will conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security.”
Microsoft’s Chief Responsible AI Officer, Natasha Crampton, framed the arrangement as a “moral necessity,” saying: “While Microsoft regularly undertakes many types of AI testing on its own, testing for national security and large-scale public safety risks must be a collaborative endeavor with governments.”
Within a week, the Commerce Department quietly deleted the webpage detailing those agreements. Anyone who clicked the old link got a simple message: “Sorry, we cannot find that page.” No explanation followed from either Commerce or the White House.
Around the same period, Commerce briefly imposed export controls on Anthropic’s newest models over concerns about potential offensive cyber capabilities, restrictions the department lifted roughly two weeks later. Anthropic said at the time that CAISI’s own researchers had “tested both our prior and new safeguards and agree that they are extraordinarily strong.”
There’s a bigger shift reportedly in the works too.
A program called Gold Eagle would hand federal officials the final say over which companies get access to Anthropic’s and OpenAI’s top-tier models, going well past the voluntary system Trump set up through his June executive order, which had simply asked firms to submit new models for government review before launch.
Featured image via X screengrab
