AI
Anthropic drafts a safety bill that could stop a model's release
Anthropic is circulating draft legislation that would let registered independent evaluators flag a model posing a foreseeable, imminent risk of killing more than 50 people or causing over $1bn of property damage, with 72 hours to notify the Commerce Department, which would then have to ask a judge to halt distribution. Labs could pay the evaluators so long as payment is not tied to the result. The thresholds mirror California's SB 53.