Demis Hassabis is done waiting for governments to figure out AI safety on their own. The Google DeepMind CEO has a concrete, controversial plan: a US-led international watchdog with actual teeth. The organization would have the authority to evaluate frontier models before they hit the market and, crucially, coordinate an industry-wide slowdown if the tech is deemed too risky. He’s not talking about a vague advisory panel.
Writing in a blog post titled “A Framework for Frontier AI and the Dawning of a New Age,” Hassabis argued the US is the only logical home for the body, citing its economic and technical dominance. He envisions something akin to the Financial Industry Regulatory Authority, staffed with independent experts and open-source community representatives. The urgency in his writing is unmistakable. AGI, he warned, “is probably only a few short years away,” describing our current moment as “the foothills of the singularity.”
This isn’t a thought experiment. According to Axios, Hassabis has spent months quietly workshopping the proposal with the Trump administration, rival AI labs, and European officials. His target is aggressive: he wants the watchdog operational before the end of 2025. Initial signals from the administration are, in his words, “very positive.” That’s a diplomatic way of saying he thinks this could actually happen.
The pitch lands at a moment of acute regulatory vacuum. The US has no comprehensive federal AI law, and global standards remain a patchwork. Hassabis, a Nobel laureate whose credibility extends beyond the boardroom, is betting that the specter of near-term AGI—and the recent push from figures like Jack Clark and Eric Schmidt to address AI’s economic shocks—will finally force a structural response. Whether sovereign nations will cede that kind of power to a US-led body is the multi-trillion-dollar question.
