Google DeepMind CEO Demis Hassabis has proposed a new independent standards body to oversee the release of frontier AI models, aiming to balance innovation with safety. The suggestion comes as the tech industry and regulators grapple with how to manage rapidly advancing AI systems.
Why a New AI Regulator Could Be a Game-Changer
In a post titled “A Framework for Frontier AI and the Dawning of a New Age,” Hassabis argues for a regulator modeled after FINRA, the financial industry’s self-regulatory organization. This body would test frontier models and develop best practices, with AI labs voluntarily sharing models for review up to 30 days before release. Over time, formal requirements could be introduced, mandating that models pass assessments before deployment in the U.S.
The proposal addresses criticism of current ad hoc reviews, such as those conducted on Anthropic’s Mythos and OpenAI’s Sol, which lacked technical expertise and transparency. Hassabis envisions the new regulator as a government-backed but industry-funded organization, staffed by technical experts and open-source representatives. It could even outsource evaluations to specialized AI safety groups.
Industry Pushback and the Path Forward
AI regulation remains contentious, with figures like White House advisor Sriram Krishnan dismissing the idea of an AI-specific regulator. Hassabis’s FINRA-inspired approach aims to sidestep these concerns by positioning the body as a self-regulatory organization. He argues that this model would keep pace with AI’s rapid advancement while incentivizing responsible behavior.
Under the proposal, the regulator could adapt to emerging risks and escalate requirements if necessary. However, its success would hinge on industry cooperation and the ability to attract top technical talent, funded by the very AI labs it oversees.
