DeepMind CEO Demis Hassabis calls for a U.S.-led body to review frontier AI models

What is Hassabis proposing?

He is proposing a U.S.-led standards body to evaluate frontier AI models before release. The organization would be industry-funded, overseen by government, and staffed in part by independent technical experts; Hassabis said labs could initially submit models for review voluntarily before any formal requirement is introduced Android Headlines,Economic Times,TechCrunch.

Why does he say this is needed now?

Hassabis argues that AI capabilities are advancing quickly and that AGI may be only a few years away, leaving a limited window to put safeguards in place. He specifically pointed to cybersecurity risks and the possibility of biological and nuclear-related threats from more capable systems Axios,NDTV,CNBC.

How would the review process work under his plan?

Hassabis proposed that frontier labs provide models to the standards body roughly 30 days before release so they can be assessed for major risks. Several reports say the process would begin as voluntary and could later become a condition for deployment in the U.S. market if the protocol proves effective Yahoo,Indian Express,TechCrunch.