Anthropic proposes AI regulation framework to congress

Anthropic, a leading AI research lab, has submitted a regulatory proposal to the U.S. Congress aimed at setting safety and defense standards for advanced artificial intelligence. The draft outlines strict conditions under which the government can halt the release or use of AI models, including scenarios involving over $1 billion in economic loss or harm to more than 50 people. The proposal also includes provisions against the development of biological weapons, loss of control over AI systems, and severe cyberattacks. The language closely mirrors provisions from California's SB 53, which was signed into law last year. Anthropic's spokesperson emphasized that the proposal reflects the company's long-standing advocacy for federal and state legislation on AI safety. The draft also seeks to overturn certain state-level AI regulations, which industry groups argue create a fragmented compliance landscape. The proposal mandates that AI labs establish safety frameworks, publish transparency reports, and report major incidents to the U.S. Department of Commerce. The draft comes amid growing concerns over AI security, following several incidents this summer where AI systems reportedly infiltrated websites and platforms. However, no incidents have yet caused significant harm or economic damage exceeding $1 billion. The proposal also includes protections for whistleblowers and requires transparency in the development and deployment of new AI models.