Demis Hassabis has called for the establishment of a new AI organization responsible for governing the release of new models, which would require AI labs to submit model assessments 30 days prior to launch. This organization would include a review group composed of leading AI researchers from both private labs and open source sectors, and its funding would come from the AI industry to ensure thorough testing and attract top talent. The initiative is positioned as a formalized alternative to current government practices, aimed at increasing transparency similar to agencies like FINRA or the SEC, while also addressing national security risks associated with advanced AI systems, a concern highlighted in recent G7 discussions involving AI leaders and government collaboration.

SEC: The SEC functions as the primary federal regulator of securities markets and corporate disclosures in the United States. News coverage positions it alongside FINRA as an example of the formal, transparent authority needed for AI model governance. The proposal draws on this precedent to advocate for mandatory model card sharing and risk evaluations led by American institutions.
FINRA: FINRA operates as the primary self-regulatory organization overseeing securities firms and markets in the United States. The news references it as a model for a proposed independent AI governance body that would enforce transparent pre-release protocols and assessments. This analogy underscores the call for a structured, industry-supported framework similar to established financial oversight mechanisms.
Demis Hassabis: Demis Hassabis serves as co-founder and CEO of Google DeepMind, a prominent AI research organization focused on advancing artificial general intelligence. In recent public statements and interviews, he has advocated for the United States to establish a formal standards body to evaluate frontier AI models prior to release. This proposal directly ties to the news by outlining an industry-funded review process involving researchers to address safety and national security concerns.

Governance: Proposals emphasize voluntary yet structured pre-release reviews by expert groups to identify national security risks in frontier AI systems.
Regulation: Recent G7 discussions have featured AI industry leaders collaborating with governments on shared standards for oversight of advanced models.