Hassabis pitched an IAEA-style AI safety body to Trump officials
Demis Hassabis went to the White House with an oversight plan before he handed DeepMind to his successor — and the details are only now coming out.
Before stepping down as DeepMind CEO, Demis Hassabis pitched a new independent industry AI safety entity, modeled on the International Atomic Energy Agency, to top Trump administration officials, according to sources cited by the Wall Street Journal. The WSJ reports he discussed the proposed body with the heads of other frontier AI labs and with administration officials including Treasury Secretary Scott Bessent. The pitch reportedly took shape around the same time as the leadership shake-up that put Koray Kavukcuoglu in charge of DeepMind — the transition we covered yesterday in Kavukcuoglu takes the DeepMind helm as Brin pushes all-in on Gemini.
The report dovetails with a public push Hassabis has been making for months. In a LinkedIn post this week he called for a federally overseen "Frontier AI Standards Body" — a public-private partnership in the mold of FINRA, the US financial self-regulator — that would "develop assessment protocols and work with appropriate federal agencies and the US National Labs to conduct testing in areas relevant to national security," including independent technical experts and open-source representatives. Fortune noted in July that the idea was "gaining momentum," with the White House reportedly receptive. The IAEA framing the WSJ reports is the more ambitious version: an independent entity with international authority over safety standards, rather than a domestic self-regulatory body.
The significance is timing as much as substance. Hassabis spent years arguing frontier AI should be governed like nuclear technology — he has repeatedly invoked the IAEA model in interviews, most recently warning that the window before AGI is "precious" and that the world is unprepared. That he carried the pitch into the White House — and to other labs — suggests the conversation about independent oversight is no longer theoretical, even in an administration that has favored deregulation. It also lands alongside reports that the White House is weighing expanded oversight of open models, which we covered earlier this week in White House expected to expand AI oversight to open models.
What to watch: whether the entity Hassabis pitched takes concrete form — and whether his successor at DeepMind picks up where he left off.
If an IAEA for AI is the right model, who should hold the inspection power — governments or an independent industry body? Tell us in the comments.
