Google’s Hassabis pushes for U.S.-led AI governance body
Google DeepMind CEO Demis Hassabis urges the US to create a new AI watchdog modeled on FINRA to test advanced models before release and coordinate slowdowns if risks emerge. He warns cyber, biological, and nuclear threats could appear within 18 months. The plan, outlined in his manifesto, would require frontier models—open or closed—to pass safety tests. Hassabis has briefed the Trump administration and other labs, aiming for the body to be operational by year-end. He says AGI is “a few short years away.”
