Demis Hassabis proposed a public-private watchdog that would screen frontier AI models before U.S. deployment, funded by the industry
JUSTIN TALLIS / Getty Images
Google $GOOGL DeepMind CEO Demis Hassabis called Tuesday for the U.S. to establish an AI standards body modeled on the Financial Industry Regulatory Authority, the private, industry-funded watchdog that polices Wall Street under Securities and Exchange Commission oversight.
Hassabis, a Nobel laureate, laid out the plan in an essay he published Tuesday on X $TWTR under the title "A Framework for Frontier AI and the Dawning of a New Age." He said "urgent action" was needed to address risks tied to artificial general intelligence, which he described as probably only a few short years away. "We've already seen the challenges frontier models pose for cybersecurity, and other threats including nuclear and bio risks may soon emerge as capabilities continue to advance," he wrote.
The proposed body would be structured as a federally overseen public-private partnership with a board composed of independent technical experts, open-source representatives, and government officials, according to CNBC. Hassabis said industry would likely foot the bill, and that the organization would need considerable financial backing to hire top-tier technical staff and run the compute-intensive evaluations required for large-scale model testing.
Initially, frontier labs would submit their models to the body ahead of release — with a window of up to 30 days — on a voluntary basis. After the regime demonstrated it was working, Hassabis envisions compliance becoming a prerequisite for any model to go live for American users. According to Axios, the framework would cover every frontier-class model — domestic or foreign, proprietary or open-weight — with eligibility thresholds updated as capabilities grow. Hassabis said he wants the body operational before year-end.
Testing would probe for dangerous cyber, biological, and deception capabilities, Hassabis said. Specific checks could include looking for attempts to bypass safety guardrails and verifying that AI-generated images carry digital watermarks.
Hassabis told Axios that the administration's handling of Anthropic's Mythos and Fable models last month — which he called "a bit of a wake-up call" — showed that Washington requires a formal governance structure rather than improvised, case-by-case responses. He said the conversations he has had with the administration have been positive.
The call comes roughly a month after Hassabis and Anthropic CEO Dario Amodei jointly urged world leaders at a closed-door G7 session in Évian-les-Bains, France, to back a U.S.-led AI governance framework. That gathering included President Donald Trump and other G7 heads of state, along with executives from several major technology companies. No concrete commitments emerged from that meeting. OpenAI CEO Sam Altman has separately called for a similar international standards body, according to CNBC.
Hassabis said that other major lab leaders agree at a high level that the industry needs to move in this direction, though Anthropic's Dario Amodei has called for a distinct FAA-style agency with the power to block unsafe models outright.
Join 500,000+ readers who start their day with Quartz.
By subscribing, you agree to our Terms of Service and Privacy Policy.
