Sources: Demis Hassabis pitched a new independent industry AI safety entity, modeled on the IAEA, to top Trump officials before stepping down as DeepMind's CEO
Demis Hassabis discussed the proposed new entity with heads of other AI labs and Trump administration officials including Scott Bessent
Context & Ripple Effects
Hassabis had already planned Washington meetings on a US-based standards body and, with Dario Amodei, backed a US-led coalition for AI rules and standards at the G7. The latest proposal gives that governance campaign a more institutional form as Hassabis leaves the DeepMind chief role after reportedly stepping back from its day-to-day work.
First-order effects
- AI lab leaders and Trump administration officials now have an IAEA-modeled proposal to assess as a potential independent safety forum.
- Hassabis’s departure from the DeepMind CEO role separates his public governance agenda more clearly from his day-to-day management of the lab.
Second-order effects
- A shared independent entity would press participating labs to align around common safety standards rather than pursue separate policy positions in Washington.
- Trump administration officials gain a single proposed channel for engaging multiple frontier AI labs on safety oversight, alongside bilateral contact with companies.
Third-order effects
- If the proposal attracts lab and government backing, AI safety governance would move toward standing institutions with cross-lab legitimacy rather than summit-based coalitions and individual company commitments.
- The effort is part of a broader contest over whether frontier-AI rules are set through US-led coordination, independent bodies, or company-led standards.
The trend: Frontier AI labs are seeking durable, state-linked governance institutions that can turn voluntary safety coordination into a shared standards system.