The US government's top AI standards job has turned over three times in four months

Key points

• Chris Fall resigned as director of the Center for AI Standards and Innovation in July 2026, roughly three months after being appointed.
• Fall is the third person to hold or vacate the role since March 2025, after Collin Burns and David Sacks.
• The White House signed an executive order in July 2026 for a new AI safety programme called "Gold Eagle" but did not name CAISI as a participant.
• In June 2026 the Commerce Department banned Anthropic's Mythos and Fable AI models under an export control order, then lifted it before the month ended.
• Google DeepMind CEO Demis Hassabis called in mid-2025 for an independent, industry-run body to set standards for advanced AI models.

The federal government's main AI standards body just lost its third leader in four months.

Chris Fall has resigned as director of CAISI, the Center for AI Standards and Innovation, which sits inside the National Institute of Standards and Technology (NIST) and is responsible for setting safety benchmarks and testing methods for AI models. No reason was given for his departure. Before running CAISI, Fall led the Department of Energy's Office of Science in the first Trump term and had also served in an acting capacity at its Advanced Research Projects Agency-Energy.

Fall had only been in the job around three months. He came in after Collin Burns left in less than a week. Burns was reportedly pushed out in April because he'd previously worked for Anthropic, a company the Trump administration was clashing with at the time, according to The Washington Post. Before Burns, the role was held by David Sacks, who served as the White House's AI and crypto policy chief before stepping down in March.

Three departures. Four months.

CAISI is the body supposed to test AI tools for safety risks and set the technical rules that companies and government agencies follow. Without stable leadership, that work stalls. The timing isn't comfortable.

In June, the Commerce Department used an obscure export control rule to effectively force Anthropic to pull its Mythos and Fable models from the market. Commerce Secretary Howard Lutnick lifted the ban by month's end, saying he was satisfied with Anthropic's safety plans. None of that involved CAISI directly.

This month, the White House signed an executive order creating "Gold Eagle", a new AI safety programme serving as a clearinghouse for cybersecurity vulnerability coordination. Several federal departments were named as participants. CAISI wasn't among them, as CNBC first noted.

Should ordinary users be worried about this leadership gap?

Not immediately, but it's worth paying attention. CAISI has published a handful of reports assessing the capabilities of Chinese open-weight models, meaning models anyone can download and run locally. The agency hasn't publicly explained how it tests these models, and as of mid-July 2026 hadn't responded to repeated press questions about its methods.

We covered the broader surge in Chinese open-weight adoption on 14 July in our piece on developer download trends, which found Chinese models taking 41% of Hugging Face downloads this spring. That context makes the CAISI vacancy feel more consequential, not less.

A new Chinese model called Kimi from lab Moonshot drew attention recently after tests showed it competing with leading Western AI models. Reports suggested the administration was weighing a ban on Chinese open models entirely, a proposal that drew criticism from Sacks himself, who argued bans shouldn't be used to shield US commercial AI companies from competition.

The agency responsible for making sense of all this has no permanent leader and a spotty public record. That's the gap that needs filling.