DeepMind CEO: FINRA-Style AI Regulation Proposed

Demis Hassabis, CEO of Google DeepMind, has proposed a new regulatory framework for artificial intelligence models. He suggests a self-regulatory body, similar to FINRA in the financial industry. This body would review frontier AI models before their release, with technical experts staffing it. AI labs would fund the system. This proposal aims to address criticisms of current ad hoc safety evaluations, which have raised questions about technical review and the transparency of recommendations. Under Hassabis's plan, the body would have a 30-day window to assess models before each release. Initially, labs would share models voluntarily. The body would be backed by the United States government but financed and independently governed by the AI industry. This approach is designed to be technically focused while supporting innovation. It aims to reduce dependence on legislative action. This proposal is important because it seeks to fill a gap in AI oversight, providing a formal review process without requiring new laws.