Hassabis pitched "IAEA for AI" to Trump officials before DeepMind exit

What happened: Demis Hassabis pitched senior Trump administration officials on a new independent AI safety body modelled on the International Atomic Energy Agency, according to the Wall Street Journal, as reported by OODA Loop.

The details: The talks, held in the weeks before he stepped down as Google DeepMind CEO on 5 August, involved Treasury Secretary Scott Bessent and science adviser Michael Kratsios, plus other frontier AI companies. The proposed body would set safety guardrails and best practices for developing artificial general intelligence.

The context: Hassabis is now DeepMind chair and Alphabet chief scientist, and has been spending more time on the oversight idea. His broader plan envisions national AI safety institutes, like the UK's, feeding into a global body that runs safety benchmarks, certification and independent audits of frontier systems.

Why it matters: The pitch follows agent incidents at OpenAI and Anthropic this summer that have pushed labs toward common safety standards. A group of labs has since begun discussing an industry-led standards body of its own.

Who's affected: Rival frontier AI developers were part of the pitch alongside government officials, suggesting Hassabis is seeking buy-in across the industry, not just from Google.

The rivals: In June, Hassabis and Anthropic's Dario Amodei jointly urged a US-led coalition on AI rules at a G7 meeting, signalling cross-company alignment on the issue.

In their words: Hassabis said earlier this year on the 20VC podcast: "Nobody should be building systems that are capable of deception because then they could be getting around other safeguards."

What's next: On 14 July, Hassabis had already called for a US-led, government-overseen public-private body to test frontier models for national-security risks, as CNBC reported. The "IAEA for AI" concept was endorsed by UN Secretary-General António Guterres in 2023, giving Hassabis's push a prior diplomatic foothold.