DeepMind CEO Pushes for Independent Frontier AI Regulator

In a recent X post, Google DeepMind CEO Demis Hassabis advocated for the establishment of a novel regulatory entity to oversee the release of advanced artificial intelligence models. Titled “A Framework for Frontier AI and the Dawning of a New Age,” his post outlined the need for a “standards body,” drawing inspiration from the Financial Industry Regulatory Authority (FINRA), which would be tasked with testing frontier models and developing best practices for their deployment.

Hassabis proposed a phased implementation, stating, “Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release.” He elaborated that “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. Labs would also work with the Standards Body to address any critical post-release vulnerabilities.” This proposed framework aims to build upon the ad hoc reviews previously conducted by the US government for models such as Anthropic’s Mythos and OpenAI’s Sol. Those earlier reviews had faced considerable criticism for their perceived lack of technical expertise and opaque decision-making processes regarding model releases. Under Hassabis’s vision, these critical decisions could be entrusted to a new, independent organization, backed by the US government but funded by the AI industry itself.

The prospect of AI regulation remains a contentious topic, sparking debate within both the tech industry and, notably, the Trump Administration. Most recently, White House AI advisor and a16z general partner Sriram Krishnan explicitly dismissed the likelihood of an AI regulator operating within the executive branch, asserting that “there will not be an FDA for AI.”

Structuring the standards body as a self-regulatory organization, akin to FINRA, could offer a viable solution to some of these concerns. Hassabis envisions the regulator staffed by representatives from the open-source community and technical experts drawn from the industry, with financial backing from AI labs crucial for retaining such talent. He also suggested the potential for outsourcing some evaluations to the expanding pool of AI safety groups, enabling specialized assessments of specific risks.

Hassabis underscored the advantages of his proposal, arguing, “The strength of this approach is it would be technically focused, while at the same time supporting innovation and incentivising responsible behaviour.” He further emphasized its adaptability, adding, “It is designed to keep up with the field’s acceleration and adapt to the biggest risks as they are identified, and could be ratcheted up if the seriousness of the situation demands.”