Google DeepMind CEO Demis Hassabis has proposed a US-led regulatory body modeled after FINRA to conduct safety screenings on frontier AI systems. The plan aims to establish a systematic framework for managing risks, including potential cyber and biological threats.