DeepMind’s CEO calls for the creation of an independent standards body to regulate frontier AI

In an X post on Tuesday morning, Google DeepMind CEO Demis Hassabis called for the creation of a new regulatory body to oversee frontier model releases. Titled “A Framework for Frontier Artificial Intelligence and the Emergence of a New Era” This publication outlines the need to establish a “standards body” modeled on the Standards Commission. Financial Industry Regulatory Authority (FINRA), which can test leading models and develop best practices for issuance.

“Initially, Frontier Labs will voluntarily share models with the standards body for review for up to 30 days before release,” the post said. “Once the evaluation protocol is proven to be effective and robust, formalization could quickly follow, meaning Frontier Models will be required to pass it to be deployed in the US market. The labs will also work with the standards body to address any critical vulnerabilities after release.”

The proposed system will build on custom reviews conducted by the US government on Anthropic’s Mythos and OpenAI’s Sol. Those reviews drew attention Great criticism Due to a lack of technical expertise and a lack of transparency in the decision-making process regarding when to release the model. Under the regulatory body proposed by Hassabis, these decisions could be handed over to a new organization, supported by the US government but funded by the AI ​​industry and operating independently.

The prospect of regulating AI remains controversial for both the tech industry and the Trump administration. and, most recently, White House AI Advisor and a16z General Partner Sriram Krishnan Reduced probability From an AI regulator within the executive branch, she said: “There will be no FDA for AI.”

Establishing a standards body as a self-regulatory organization like FINRA could be a way to address these concerns. Hassabis envisions the regulatory body being staffed with open source representatives and technical experts from within the industry, along with the financial support from AI labs that will be necessary to retain them. They could also outsource some assessments to a growing group of AI safety groups that will be able to specialize in specific risks.

“The strength of this approach is that it will focus on the technical aspect, while at the same time supporting innovation and incentivizing responsible behavior,” Hassabis says. “It is designed to keep pace with the industry’s acceleration and adapt to the greatest risks identified, and can be escalated if the severity of the situation requires.”