DeepMind CEO proposes independent AI standards body

Google DeepMind CEO Demis Hassabis has proposed the establishment of a new, independent regulatory body to oversee the development and release of "frontier" artificial intelligence models. Hassabis outlined the idea in an X post on Tuesday, titled "A Framework for Frontier AI and the Dawning of a New Age."

The proposed organization would function similarly to the Financial Industry Regulatory Authority (FINRA) in the finance sector. Its mandate would involve testing advanced AI models and developing best practices for their deployment. Initially, "frontier AI labs" would voluntarily submit their models for review up to 30 days prior to public release.

Hassabis envisions that once the assessment protocol proves effective and robust, formalization could follow, mandating that frontier models pass this review to be deployed in the U.S. market. The body would also collaborate with labs to address critical post-release vulnerabilities. This approach aims to improve upon past ad hoc government reviews, which faced criticism for a lack of technical expertise and opaque decision-making.

The proposal faces industry skepticism, with a White House AI advisor having previously dismissed the idea of an "FDA for AI." Establishing the body as a self-regulatory organization, akin to FINRA, could potentially navigate these concerns. Hassabis suggests the organization would be staffed by open-source representatives and technical experts from within the industry, funded by AI labs.