Head of America’s AI safety agency resigns after three months
July 21, 2026 – 8:17 am
Image by: The White House
Chris Fall’s exit leaves the Center for AI Standards and Innovation (CAISI) without a permanent leader as Washington rethinks how it polices frontier models.
The director of the United States’ main AI oversight body has resigned after roughly three months in the job, leaving the agency without permanent leadership at a delicate moment for federal technology policy.
Chris Fall has stepped down as head of the Center for AI Standards and Innovation (CAISI), the Commerce Department unit housed within the National Institute of Standards and Technology, in a departure confirmed on Monday.
CAISI is the body formerly known as the US AI Safety Institute, which the Trump administration renamed and recast earlier in its term, shifting the emphasis from safety towards standards and innovation. The rebrand came alongside a wider effort to loosen federal oversight, including an executive order that asks AI developers to submit new models for voluntary review rather than mandatory testing.
The heart of EU tech news, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
Renaming was more than cosmetic. Dropping the word "safety" signalled a preference for light-touch, innovation-friendly rules, and it lined the agency with an administration that has been sceptical of binding constraints on how AI gets built.
The Commerce Department confirmed the resignation without giving a reason, and an official said Fall’s appointment had been “always intended to be temporary”. He took up the post in late April, which makes his tenure one of the shortest at the top of the agency.
NIST director Dr Arvind Raman will oversee CAISI and serve as acting director in the interim, according to the department, which said it expects to name a permanent successor within weeks. Raman only took charge of NIST at the end of June, so the arrangement folds two demanding briefs into a single office.
Fall’s exit extends a run of turnover across Washington’s AI ranks. His predecessor, Collin Burns, was pushed out after four days over his ties to Anthropic, and David Sacks, the White House AI and crypto czar, left in March without a named replacement. Neither departure came with a detailed public explanation, feeding a sense that AI policy roles in this administration have become unusually precarious.
That churn has coincided with an internal fight over AI oversight that has repeatedly stalled efforts to formalise the government’s testing role. CAISI sits at the centre of that work, examining unreleased systems for cybersecurity weaknesses and national security risks. The unit collaborates with leading labs, including Anthropic, Google DeepMind, and OpenAI, and it underpins the pre-release evaluations that several developers have agreed to despite the absence of any statutory mandate.
Its authority rests on voluntary cooperation rather than law, which leaves the office reliant on goodwill from the companies it is meant to scrutinise. The agency’s public posture has been erratic under the current administration. In May, the Commerce Department quietly deleted a webpage that had detailed a security-testing agreement with Microsoft, Google, and xAI, a move it never explained publicly.