Member-only story
Anthropic’s Bill Kicks In at 50 Deaths and $1bn. It Also Switches Off Three State Laws.
The company wrote the rules, chose the number that triggers them, and pays the people who check its homework. Congress is being asked to make that the floor.
On 16 October, Politico published a story by Owen Dahlkamp about a confidential draft bill that Anthropic has been circulating on Capitol Hill. The outlet obtained the draft itself, and nobody had reported it before. The company has a line for it: “This bill represents the proposals outlined in Anthropic’s Advanced AI Framework… These are sensible policies we believe Congress should act on.”
Fine. Here is what these so called sensible policies do.
Under the draft, the federal government could block a model’s release or use only where it finds a foreseeable, imminent risk of more than $1bn in damage, or of killing or seriously injuring more than 50 people. The risk has to come from one of three sources: a bioweapon, a cyberattack, or a model escaping its developer’s control. Only the Commerce Department could petition a federal court for that order, and the text bars every other agency and official from suspending a model. Independent evaluators would register with Commerce, could be paid…
