Google DeepMind's Boss Wants a Global AI Watchdog Before the Technology Gets Away From Us
Demis Hassabis has spent months briefing governments and rival labs on a plan for an independent body that could pause AI development if things go wrong. He wants it running before 2026.
Key points
- Google DeepMind CEO Demis Hassabis published a blog post calling for a new international body to oversee the most powerful AI systems before they are released.
- Hassabis argues the United States should lead the effort, given its economic and technical standing as the natural home for setting global AI standards.
- The proposed watchdog would have real authority: it could evaluate the most powerful AI models before launch and coordinate an industry-wide slowdown if a model was judged too dangerous to release.
- Hassabis, joint winner of the 2024 Nobel Prize in Chemistry, has spent months briefing the Trump administration, rival AI labs and European officials on the plan.
- No binding global AI framework exists, and no comprehensive national AI law exists in the United States.
Demis Hassabis runs Google DeepMind, the AI research division behind some of the most advanced artificial intelligence systems in the world. He also just shared what may be his most ambitious proposal yet: a powerful, independent watchdog that could put the brakes on AI development if things start to look dangerous.
In a blog post titled "A Framework for Frontier AI and the Dawning of a New Age," Hassabis argued that AI regulation is becoming urgent, not optional. Frontier models, meaning the most capable AI systems that push the boundaries of what the technology can do, are advancing fast. Too fast, he suggests, for the world to stay passive.
His proposed body would work a bit like FINRA, the Financial Industry Regulatory Authority that polices financial firms in the United States. It would be staffed by independent experts and representatives from open-source AI communities, groups that build and share AI tools freely. It would also carry real power: reviewing the most powerful models before release and calling a halt across the whole industry if a model looked too risky to deploy.
First reported by The Verge, citing Axios, Hassabis has spent months quietly building support behind the scenes, briefing the Trump administration, rival labs and European officials. His read on the White House response: "The noises I've been hearing are very positive," he told Axios. He wants the organisation running before the end of this year.
Why the urgency? Hassabis put it plainly. Artificial general intelligence, or AGI, meaning an AI system capable of matching or outperforming human beings at most intellectual tasks, is "probably only a few short years away."
He's not alone in sounding the alarm. Anthropic cofounder Jack Clark and former Google CEO Eric Schmidt recently joined economists urging world leaders to take the economic disruption AI could cause far more seriously. Our story on the Australian Labor MP pushing for tougher AI rules shows this pressure is building across governments, not just in Washington.
For ordinary people the stakes are concrete. No binding global rules on AI exist, and no national law covers the ground in the United States. Hassabis is trying to change that before the technology outpaces the conversation.
Should the rest of us be worried about the gap in AI rules?
Not panicked, but paying attention. Companies currently set many of their own limits. An independent watchdog with real authority would shift that balance toward public accountability. Whether the Trump administration's warm words translate into action before 2026 is the thing to watch.
