Demis Hassabis Calls for US Frontier AI Standards Body for National Security Testing

Google DeepMind CEO Demis Hassabis advocates for a US Frontier AI Standards Body to conduct national security-relevant testing on advanced AI models. This body would initially offer voluntary 30-day reviews for "Frontier Models" from "Frontier Labs" before making assessments mandatory for US market deployment. The proposal also encourages Frontier Labs to adopt best practices like publishing model cards and enhancing cybersecurity.