Demis Hassabis Calls for US-Led Global AI Watchdog
Google DeepMind CEO Demis Hassabis is advocating for a global AI regulatory body with teeth — one that could halt dangerous AI development — and wants the United States to lead it.
Original sourceIn a notable shift for a sitting Big Tech AI executive, Demis Hassabis publicly called for an international AI oversight organization modeled loosely on existing global regulatory bodies. Speaking to The Verge, Hassabis argued that the pace of AI development has outrun the ability of any single nation to govern it effectively, and that a coordinated global mechanism — with actual authority to pause or stop dangerous projects — is now necessary.
Hassabis specifically named the United States as the right country to anchor such an organization, citing its current dominance in frontier AI research and its existing institutional relationships with allied governments. The proposal is notable not just for its content but for its source: Hassabis runs one of the most advanced AI labs on the planet, and a watchdog with halt authority would, in theory, apply directly to Google DeepMind's own work.
The call comes amid a fragmented global policy landscape. The EU AI Act is in various stages of enforcement, the US has oscillated between executive orders and congressional inaction, and China has pursued its own regulatory path largely outside Western frameworks. A US-led watchdog would face immediate questions about jurisdiction, enforcement, and whether it could compel cooperation from non-allied states developing frontier models.
Critics will note the inherent tension in a lab executive proposing regulation that — if designed as described — would likely entrench established players while raising barriers for newer entrants. Whether this is genuine safety advocacy, strategic incumbency protection, or both is a question the proposal itself does not resolve.
Panel Takes
The Skeptic
Reality Check
“The oldest move in the incumbent playbook: propose regulation that sounds universal but structurally advantages whoever helped design it. A US-led watchdog with halt authority sounds serious until you ask who defines 'dangerous,' who sits on the board, and whether it has any mechanism to touch state-sponsored labs in countries that won't sign on. Hassabis gets credit for saying it publicly, but a proposal without an enforcement mechanism for non-participants isn't safety policy — it's a moat with extra steps.”
The Futurist
Big Picture
“The thesis here is that frontier AI risk is a coordination problem, not a knowledge problem — that we know enough to govern, we just lack the institution to do it. That's a falsifiable bet, and the dependency is stark: it only works if the US actually wants to lead multilaterally rather than use AI dominance as a unilateral geopolitical lever, which recent history does not strongly support. The second-order effect nobody is talking about is what a functioning global AI watchdog does to the innovation geography — labs outside the US-aligned bloc don't slow down, they just operate outside the regime, which may make the risk landscape worse, not better.”
The Founder
Business & Market
“Read the incentive structure before you read the proposal. Hassabis is the CEO of a frontier lab backed by one of the largest companies on earth — a regulatory body with halt authority and US leadership is a filter that Google DeepMind is almost certainly positioned to pass, while the next generation of well-funded challengers might not be. That's not a conspiracy, it's just how incumbents rationally engage with regulatory design. The tell will be whether DeepMind's proposed framework includes independent technical evaluation or relies on self-reporting — one of those is a watchdog, the other is theater.”
The PM
Product Strategy
“The job-to-be-done here is 'prevent catastrophic AI development outcomes' and the proposed solution is a new international institution — which is a product with a notoriously long time-to-value and a brutal onboarding problem involving sovereign governments. The proposal doesn't yet answer the one question that determines whether it's a real product or a concept: what does the watchdog actually do on day one, before it has the political capital to halt anything? A credible roadmap from 'new body is formed' to 'dangerous project is actually stopped' is the gap between a press release and a policy.”