ANTHROPIC QUIETLY PITCHES ITS OWN AI SAFETY BILL TO LAWMAKERS

Summary from the TLDR InfoSec newsletter

Representatives from Anthropic have quietly pitched a draft AI safety bill to lawmakers. Under the bill, the government could only block a model's release or use if it poses a foreseeable, imminent risk of more than $1B in damages, or of killing or seriously injuring more than 50 people via the creation or release of a biological weapon, the model evading its developers' controls, or a cyberattack. Independent evaluators would register with the Department of Commerce and report any such risk within 72 hours.