Anthropic is taking its AI safety pitch to Congress. The artificial intelligence company has confirmed it is sharing a draft bill with U.S. lawmakers that would set a deliberately high bar for government intervention: a model could be blocked only if it poses an imminent risk of killing or seriously injuring more than 50 people, or causing more than $1 billion in damages.
Under the draft, that risk would have to come from one of three sources — the creation or release of a biological weapon, a model slipping out of its developer's control, or a cyberattack. The process would run through a judge and the Commerce Department, with independent evaluators registered with Commerce. The bill has not been formally introduced, and no sponsor had been publicly identified as of this weekend.
The draft would also require frontier AI labs to test their models, publicly disclose their findings, submit to independent evaluation and maintain strong security programs — requirements drawn from Anthropic's “Advanced AI Framework,” which the company has promoted since June.
The proposal lands in the middle of a widening fight over how AI should be governed: state legislatures are writing their own rules, the industry is split between disclosure-first and capability-first camps, and a threshold written around catastrophic harm would leave most commercial AI products untouched while putting the most powerful systems under a microscope.
