CAISI's Third Director Quits in Months, Leaving AI Safety Vacuum

The Center for AI Standards and Innovation (CAISI) is now hunting for its fourth leader in roughly four months. Chris Fall, appointed just three months ago, has resigned, multiple outlets confirm. No reason was given for his departure, but the move extends a staggering leadership churn that began when the agency’s inaugural chief, Collin Burns, was reportedly pushed out in under a week due to his former employment at Anthropic. The White House’s AI and crypto czar David Sacks had preceded Burns as the de facto head before stepping down in March.

Fall’s exit leaves a void at the very moment the government is scrambling to define its AI safety architecture. The agency, housed under NIST, is supposed to be the tip of the spear for technical standards and model testing. Yet it was notably absent from the White House’s new ‘Gold Eagle’ cybersecurity initiative earlier this month, an executive order that named a slew of federal organizations but conspicuously left CAISI off the roster. The silence from the Department of Commerce and NIST regarding how CAISI actually evaluates large language models is deafening; TechCrunch reports its inquiries since July 9 have gone completely unanswered.

The instability at CAISI mirrors a broader, messier struggle over who gets to regulate frontier AI. The vacuum is already attracting outside players. After a brief, bizarre ban on Anthropic’s models in June—triggered by an obscure export rule and lifted after the Commerce Secretary personally approved safety plans—Google DeepMind CEO Demis Hassabis publicly called for an independent, industry-run standards body modeled after FINRA. That’s a direct challenge to the whole premise of a government-led CAISI. The situation grew more fraught this weekend as the administration reportedly considered banning Chinese open models like Moonshot’s Kimi, sparking immediate backlash from Sacks himself, who warned against using regulations as a protectionist cudgel for domestic proprietary labs.

CAISI has managed to publish a few reports on Chinese open-weight models like GLM-5.2 and DeepSeek V4 Pro, but its testing processes remain a black box. With no permanent leadership, a mission being encroached upon by private-sector proposals, and a White House that seems to route critical safety initiatives around it, the agency’s relevance is increasingly in question. One has to wonder if the next director will get a real mandate—or just become the next name on a very short, very chaotic list.