US AI Safety Agency Head Resigns After 3 Months

The Head of America's AI Safety Agency Resigns After Three Months — the Second Exit This Year

Chris Fall is out at CAISI, whose predecessor lasted less than a week. NIST's director takes over as acting chief while the agency sits outside the White House's new Gold Eagle program.

Jul 21, 2026Source: TechCrunch

The revolving door at America's AI model-testing agency keeps spinning. Dr. Chris Fall has resigned as director of the Center for AI Standards and Innovation (CAISI) — the Commerce Department body housed under NIST and formerly known as the US AI Safety Institute — the department confirmed Monday, July 20, first reported by Axios. He lasted roughly three months.

A pattern, not an incident

Fall's exit is the second abrupt departure at CAISI this year. His predecessor, Collin Burns, left in April after less than a week in the job — reportedly pushed out over his prior work at Anthropic amid the administration's friction with that company. Before CAISI, Fall ran the Department of Energy's Office of Science in the first Trump administration and served as acting director of ARPA-E — a résumé that made him a conventional, credentialed pick.

No reason given

Commerce offered no explanation for the departure, with an official saying the appointment was always intended to be temporary. Spokesperson Kristen Eichamer said NIST director Dr. Arvind Raman "will continue to oversee CAISI and will serve as Acting CAISI Director," with a permanent director expected "in the coming weeks." Raman has reportedly been vetting candidates for weeks — suggesting the transition was in motion before Monday's confirmation.

An agency adrift

The churn lands at a delicate moment. CAISI is nominally the US government's technical arm for evaluating frontier models, but it was notably absent from the White House's recent Gold Eagle AI oversight program, which now shapes who gets access to frontier systems. David Sacks stepped down as White House AI and crypto czar in March, adding to the sense that AI policy authority in Washington is concentrated in fewer, more political hands.

What to watch

Whether the permanent pick is a technical evaluator or a political appointee will signal what CAISI becomes: a real testing body with teeth, or a vestigial letterhead while frontier-model oversight migrates to the White House.

Why It Matters

Two directors gone in four months means the US effectively has no stable technical authority for frontier-model evaluation — while the EU staffs up its AI Office and China institutionalizes governance through WAICO. For labs, the practical oversight now runs through the White House's Gold Eagle program, making model access a political relationship rather than a standards process.