Google DeepMind's Boss Wants a Global AI Watchdog Before the Technology Gets Away From Us

Key points

• Google DeepMind CEO Demis Hassabis published a blog post in 2025 calling for a new international body to oversee the most powerful AI systems before they are released to the public.
• Hassabis argues the United States should lead the effort, citing its economic and technical standing as the natural home for setting global AI standards.
• The proposed watchdog would have real authority: it could evaluate cutting-edge AI models before launch and coordinate an industry-wide slowdown if a model was judged too dangerous to release.
• Hassabis, a joint winner of the 2024 Nobel Prize in Chemistry, has spent months briefing the Trump administration, rival AI companies and European officials on the plan.
• No comprehensive national AI law exists in the United States today, and no binding global framework exists either.

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 and cutting-edge 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). Most importantly, it would have teeth: the power to review the most powerful AI models before companies release them, and to call a halt across the whole industry if a model looked too risky to put into the world.

First reported by The Verge AI, citing Axios, Hassabis has spent months quietly building support behind the scenes. He briefed the Trump administration, rival labs and European officials. His read on Washington? "The noises I've been hearing are very positive," he told Axios. He wants the organisation up and running before 2025 is out.

Why the urgency? Hassabis put it plainly in the blog post. 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 is 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.

For ordinary people, the practical takeaway is simple. There are currently no binding global rules on AI and no comprehensive national law 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. The absence of an agreed framework means companies currently set many of their own limits. An independent watchdog with real authority would shift that balance toward public accountability, which is a meaningful change worth following.