DeepMind CEO calls for FINRA-style AI standards body to regulate frontier models

In a Tuesday X post, Google DeepMind CEO Demis Hassabis introduced “A Framework for Frontier AI and the Dawning of a New Age,” calling for an independent standards body to regulate frontier AI model releases. The proposed regulator is modeled after the Financial Industry Regulatory Authority (FINRA), a self-regulatory organization. Initially, frontier labs would voluntarily share models with the body for review up to 30 days before release. Once the assessment protocol proves effective, formalization would follow, requiring models to pass testing before US deployment. Labs would also work with the body to address post-release vulnerabilities.

Hassabis’s proposal follows criticism of ad hoc US government reviews of Anthropic’s Mythos and OpenAI’s Sol, which lacked technical expertise and transparency. The new body would be funded by the AI industry but operated independently, backed by the US government. Staff would include open source representatives and technical experts, potentially outsourcing evaluations to specialized AI safety groups. The plan addresses concerns about “FDA for AI” regulation, which White House advisor Sriram Krishnan recently dismissed. Hassabis argues this approach is technically focused, supports innovation, and can adapt to emerging risks.