Anthropic has presented a draft artificial intelligence safety bill to United States lawmakers that would establish conditions for restricting advanced artificial intelligence models and override certain state-level regulations. According to a report by POLITICO, the proposal would allow the government to block a model’s release or use only when it presents a foreseeable and imminent risk involving more than $1 billion in damages or serious harm to more than 50 people under specified circumstances. The draft also proposes independent evaluations of advanced models, safety frameworks, transparency reports, whistleblower protections and mandatory reporting of critical safety incidents to the Commerce Department. Evaluators would have to report identified imminent risks within 72 hours of completing their assessments. A federal judge could order a suspension following a government request, while the proposal would limit other government authorities from independently halting development or deployment. The bill would also override certain state requirements concerning risk disclosures, independent evaluations and incident reporting. Supporters of national standards argue that a unified framework could reduce compliance difficulties, while some lawmakers have raised concerns about weakening existing state protections.
The proposal comes as Congress considers different approaches to managing advanced artificial intelligence risks, government oversight and industry accountability.
Read more here
https://www.politico.com/news/2026/10/16/anthropic-ai-safety-01124654
