Google DeepMind chief Demis Hassabis is calling for a U.S.-led global AI watchdog that could review the most powerful frontier models before release and coordinate a slowdown if they are judged too risky.
Hassabis, the cofounder and CEO of Google DeepMind, argued that Washington is best placed to lead the effort, citing the country’s “economic and technical standing.” The proposed body would include independent experts and representatives from open-source communities.
Under the framework described by Hassabis, the watchdog would evaluate frontier AI systems before deployment rather than leaving safety calls entirely to individual labs. It would also have the ability to coordinate an industry-wide pause if a model appeared too dangerous to release.
The push comes as AI companies race to build systems that can handle more complex work across coding, science, search, business software and consumer apps. Hassabis said artificial general intelligence is “probably only a few short years away,” a timeline that has sharpened arguments over who should set rules for the technology.
Hassabis has reportedly spent months building support for the plan, including conversations with the Trump administration, other AI labs and European officials. He hopes the organization can begin operating before the end of the year.
The proposal adds pressure on governments that have struggled to keep pace with AI development. The United States still has no comprehensive national AI law, while global rules remain fragmented across regions and regulators.
Hassabis has also backed tougher protections against AI-assisted bioweapons production. His latest proposal leaves one question hanging over the industry: whether the companies building frontier AI will accept an outside referee before the next generation of models arrives.
