Anthropic has privately pitched lawmakers on a draft federal AI safety bill that would narrow when the government can stop a frontier model’s release, according to Owen Dahlkamp at Politico. A block would require an imminent risk of over $1 billion in damage or over 50 deaths or serious injuries, caused by a bioweapon, a cyberattack, or a model escaping its developer’s control. Only a federal judge in Washington, D.C. could suspend a model at the Commerce Department’s request, and no other agency could. The draft would also override some state rules on risk disclosure, outside evaluations, and incident reporting.
