DeepMind CEO proposes U.S.-led AI standards body for frontier models
Hassabis wants a U.S.-led AI watchdog that could screen frontier models before release and even slow launches if risks rise, raising the stakes for who writes the rules.
Demis Hassabis called for a U.S.-led standards body to screen frontier AI models before they are released, pressing for an independent watchdog that would test the most advanced systems for cybersecurity, biological and nuclear risks. The Google DeepMind chief said the group should be staffed by technical experts, funded by industry and answerable to the U.S. government, giving developers a formal role in the safety regime that could govern their own products.
Hassabis modeled the proposal on FINRA, the private self-regulatory body created in 2007 from the consolidation of NASD and parts of the New York Stock Exchange’s regulation division. FINRA oversees U.S. broker-dealers under Securities and Exchange Commission oversight, and that mix of industry funding and government supervision is the template Hassabis is invoking for frontier AI. A similar structure could be used to set best practices for release decisions and, if necessary, coordinate an industry-wide slowdown when risks mount.
The Frontier Model Forum was launched in July 2023 by Anthropic, Google, Microsoft and OpenAI to promote safe and responsible development, advance technical evaluations and share best practices. Google has its own Frontier Safety Framework, and it published an updated Responsible AI report in 2025. NIST has also issued an AI Risk Management Framework and a generative AI profile. In 2026, the largest AI companies have weakened some safety commitments even while model capabilities have advanced.
Hassabis said AGI may arrive within only a few years, and he hoped the proposed standards body could be operating before the end of 2026, with the idea receiving a positive response from the Trump administration.
This article was produced by Prism’s automated news system from verified source data, official records, and press releases, then run through automated quality and moderation checks before publishing. The system is built and supervised by the people who set the standards it runs under. Read our full AI policy.
Did this article answer your question?
