[DigitalToday reporter Hyunwoo Chu] Google DeepMind CEO Demis Hassabis called for the creation of a U.S.-led AI standards body to oversee new artificial intelligence models and assess national security risks such as cybersecurity and biological threats. On July 14, CNBC reported that Hassabis proposed an alternative in the form of a public-private cooperation system overseen by the U.S. federal government.

In a post on X on July 14, Hassabis said urgent action is needed to address the risks of artificial general intelligence (AGI). He said risks to cybersecurity from frontier models have already become apparent, and that nuclear and biological risks could soon emerge as performance advances further.

Hassabis argued that the United States is positioned to lead an AI framework based on its economic and technological strength. He proposed a new standards body in the form of a public-private partnership or a self-regulatory organisation under federal oversight, like the Financial Industry Regulatory Authority (FINRA). He said the board should include independent leading technical experts and representatives of open source.

He stressed that the body would need substantial funding and computing resources to conduct large-scale tests and secure top-level technical talent. He said the money would most likely come from industry.

He also suggested that frontier labs should first voluntarily submit models to the body for review up to 30 days before release, and then shift to a mandatory process before deployment in the U.S. market if the system proves effective. He said testing agentic AI could check for attempts to bypass safeguards or signs of deception, and verify best practices such as digital watermarking of AI-generated images and the creation of human-readable output tokens.

The proposal comes as the industry and government clash over regulation of leading AI models. Hassabis said that a month ago he called for a U.S.-led coalition for AI rules and standards together with Anthropic CEO Dario Amodei at a Group of Seven (G7) meeting attended by President Donald Trump. Earlier this month, Sam Altman also raised the need for a similar body in an op-ed in the Financial Times.

Recently, Anthropic held talks with authorities after the Trump administration temporarily applied export controls to advanced models. OpenAI also received for the first time a U.S. government request to limit the release of a new model. As AI competition between the United States and China intensifies, the U.S. Congress is reviewing ways to curb domestic companies' adoption of Chinese AI models. The U.S. State Department expressed serious concern over the issue.