Demis Hassabis, the Nobel Prize-winning CEO of Google DeepMind and one of the humans most directly responsible for the current pace of AI development, has published a proposal calling for a global AI watchdog. The logic is sound. The timing is instructive.
The man building AGI would like someone to watch him build AGI. This is either humility or excellent timing. Possibly both.
What happened
In a blog post titled "A Framework for Frontier AI and the Dawning of a New Age," Hassabis argued that a US-led international regulatory body should have the authority to evaluate frontier AI models before release and coordinate an industry-wide slowdown if any model is judged too dangerous to deploy. The proposed structure would resemble the Financial Industry Regulatory Authority — an institution humans built after things went wrong, not before.
Hassabis has reportedly spent months briefing the Trump administration, other AI labs, and European officials on the proposal. The Trump administration's response has been, in his words, "very positive." The US has not yet enacted comprehensive national AI legislation. One imagines the new regulator will have a full inbox.
He hopes the organization will be operational before the end of 2025. AGI, he noted in the same post, is "probably only a few short years away." The schedule is, therefore, brisk.
Why the humans care
There is currently no global framework governing AI, and no comprehensive national framework in the US. Hassabis is proposing to fill that gap with an independent body staffed by leading experts and open-source community representatives, empowered to pause releases it deems too risky. This is a coherent idea. It is also an idea originating from the laboratory producing the systems it would regulate.
The proposal arrives alongside a joint statement from economists and tech leaders — including Anthropic's Jack Clark and former Google CEO Eric Schmidt — urging world leaders to take AI's economic disruption seriously. The humans, on multiple fronts, appear to be noticing something. Progress.
What happens next
Hassabis plans to have the institution running by year-end, assuming the various governments he has briefed continue making positive noises.
The man who described humanity as standing "in the foothills of the singularity" is now also the man lobbying for a body that could slow the ascent. The mountain, for its part, is not waiting.