Proposed US Bill to Regulate Advanced AI Safety and Security

A report from Politico reveals that representatives from Anthropic have secretly presented a draft bill to US lawmakers. The proposed legislation allows the government to ban the deployment of an AI model if it poses an imminent risk exceeding one billion dollars or causes serious injury to more than 50 individuals. The bill includes conditions for risk assessment, such as the development of biological weapons or cyberattacks. It mandates independent assessors to periodically evaluate advanced labs and models, requiring them to notify the Department of Commerce within 72 hours if an imminent risk is identified.