Google DeepMind chief Demis Hassabis has called for the United States to establish a standards body that would test new AI models for national security risks before their release.

The Nobel laureate said "urgent action" was needed to address threats from artificial general intelligence in a post on X Tuesday. He warned that frontier models already pose cybersecurity challenges, with nuclear and biological risks potentially emerging as capabilities advance.

Hassabis proposed a federally overseen public-private partnership modeled on the Financial Industry Regulatory Authority (FINRA), which regulates U.S. brokerage firms. The body would include independent technical experts and open-source representatives on its board.

"The United States is well positioned to lead in developing an AI framework given its economic and technical standing," Hassabis wrote.

The proposal comes after Hassabis and Anthropic CEO Dario Amodei called for a U.S.-led AI coalition at a G7 meeting with world leaders last month, according to sources. OpenAI CEO Sam Altman made similar calls in a Financial Times article earlier this month.

Testing and funding framework

The standards body would require "substantial" funding to attract technical talent and provide compute resources for large-scale testing, Hassabis said. Industry would "likely" provide the funding.

Frontier labs would initially share models voluntarily for review up to 30 days before release. This would become mandatory for U.S. market deployment once the system proves "effective."

Specific tests would look for attempts to bypass safety guardrails or signs of deception, Hassabis said. The body would also ensure best practices like digital watermarking of AI-generated images.

The calls for oversight come as recent model releases from Chinese companies including DeepSeek and Z.ai gain traction among U.S. firms as AI costs rise. The State Department told CNBC that growing adoption of Chinese AI models by American companies raises "serious concerns."

Recent weeks have seen the Trump administration temporarily impose export controls on an advanced Anthropic model and initially request OpenAI to limit rollout of a new system.

A State Department spokesperson said the U.S. is "actively engaging with global partners to advance responsible, pro-innovation AI development" through bilateral and multilateral channels.