DeepMind CEO Proposes Independent Standards Body for Frontier AI Oversight

Google DeepMind CEO Demis Hassabis has proposed creating an independent standards body to evaluate frontier artificial intelligence models before they are released, arguing that a dedicated organization could provide more consistent oversight as AI systems become increasingly powerful. The proposal was outlined in a post on X titled “A Framework for Frontier AI and the Dawning of a New Age.”

Hassabis suggested a self-regulatory organization modeled after the Financial Industry Regulatory Authority (FINRA), which oversees the U.S. securities industry. Under the proposal, frontier AI developers would voluntarily submit new models for review up to 30 days before release. If the evaluation process proves effective, passing the review could eventually become a requirement for deploying frontier AI models in the U.S. market.

"Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release," Hassabis wrote. "Once the assessment protocol is shown to be effective and robust, formalisation could quickly follow, meaning that Frontier Models would be required to pass it to be deployed in the US market."

The proposed organization would also work with AI developers to address significant vulnerabilities discovered after models are released. Hassabis said the regulator should operate independently while receiving financial support from AI companies and include technical experts, open-source representatives, and specialists capable of evaluating emerging AI risks.

The proposal follows criticism of recent U.S. government reviews of Anthropic’s Mythos and OpenAI’s Sol models, which some observers said lacked sufficient technical expertise and transparency. Hassabis argued that an independent body focused on technical assessments could provide a more structured approach to evaluating frontier AI systems.

The idea arrives as debate over AI regulation continues across the technology industry and within the Trump administration. White House AI adviser and Andreessen Horowitz general partner Sriram Krishnan recently said there would not be an “FDA for AI,” highlighting ongoing differences over how advanced AI models should be governed.

Hassabis said a FINRA-style organization could help balance innovation with responsible development while adapting its evaluation methods as AI capabilities evolve. He also suggested the body could partner with specialised AI safety organisations to conduct certain technical assessments.

"The strength of this approach is it would be technically focused, while at the same time supporting innovation and incentivising responsible behaviour," Hassabis wrote. "It is designed to keep up with the field's acceleration and adapt to the biggest risks as they are identified."