Hassabis Proposes Industry-Funded US Self-Regulatory Body for Frontier AI Pre-Deployment Review
Google DeepMind CEO Demis Hassabis published a detailed proposal on July 14, 2026 for a FINRA-style self-regulatory organization that would require frontier AI labs to submit models for safety evaluation up to 30 days before release, funded by industry assessments rather than congressional appropriations. The plan is structurally lighter than competing proposals from Anthropic's Dario Amodei, who wants an FAA-style federal agency, and OpenAI's Sam Altman, who has called for an international forum, and it deliberately avoids creating a new government body to navigate the Trump administration's stated opposition to an FDA-equivalent for AI. Three design questions remain unresolved: how 'frontier' will be defined, who decides when voluntary review becomes mandatory, and how the body would relate to the Commerce Department's existing export-control authority.