DeepMind CEO: AI Watchdog Could Pause Industry Development

Google DeepMind CEO Demis Hassabis is calling for a new "Frontier AI Standards Body." This body would be modeled on the financial industry's watchdog and could, if needed, slow down AI development across the entire industry. What's interesting is this proposal comes as Washington has already informally regulated AI model releases. For example, one administration froze advanced AI models for 18 days due to national security concerns. Another AI company accepted a restricted launch of its most capable model at the government's request. Hassabis's plan aims to replace this informal gatekeeping with a structured, expert-staffed, and industry-funded institution that answers to the federal government. AI labs would voluntarily share their models with this body for safety testing up to 30 days before release. The testing would look for dangerous cyber, biological, and "deception" capabilities. The proposed board would be mostly independent, with experts and representatives from industry, government, and open-source communities. Industry would fund this body to ensure it has the resources to attract top talent and conduct large-scale testing. This could impact how future AI models are developed and released.