Google DeepMind CEO Wants the US to Build a Global AI Watchdog — With a 'Brake Pedal'

What happens when a frontier AI model starts doing something its creators didn’t anticipate? Demis Hassabis, the co-founder and CEO of Google DeepMind, wants a global body in place that can answer that question — by hitting the brakes.

Hassabis has been quietly lobbying on multiple fronts for months — the US government, other AI labs, even European officials. His proposal: a global AI oversight agency led by the United States, modeled after the Financial Industry Regulatory Authority, better known as FINRA — the self-regulatory organization that oversees brokerage firms in America.

The envisioned body would bring together independent experts and open-source community representatives to evaluate frontier AI models before they’re released. If a model poses unacceptable risks, the agency could trigger a coordinated industry slowdown — a collective brake pedal for the entire field.

Hassabis reportedly wants the agency operational by the end of this year. The timing reflects a growing unease inside the AI industry itself: the people building these systems want a mechanism to stop them when things go wrong, and they don’t want that decision left to any single company or government.

A FINRA-style structure is an interesting choice. FINRA isn’t a government agency — it’s a private-sector regulator authorized by Congress. That gives it more flexibility and technical expertise than a traditional bureaucracy, while still carrying enforcement authority. Whether other AI labs — especially those outside the US — would submit to such oversight is an open question.

For now, Hassabis is making the rounds. That the CEO of the world’s most prominent AI lab is the one pushing for external oversight says something about where the industry thinks this is headed.