Anthropic quietly pitches its own AI safety bill to lawmakers - POLITICO

Anthropic has quietly pitched Capitol Hill lawmakers on a confidential draft bill to regulate frontier AI safety and security, POLITICO reported on 16 October. The article cites three people with knowledge of the efforts. According to those accounts, the company's approach has been quiet and the draft has not been made public, with the reporting resting on descriptions of the proposal rather than on released text.

As described by POLITICO, the bill would let the government block a model's release or its use only if the model poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. The article states that the risk would have to stem from one of three causes: a biological weapon, the model evading its developer's controls, or a cyberattack. Both the scale of harm and its source are set out in the reporting as limits on when the government could act.

The article describes the pitch as quiet and the draft as confidential, and attributes its account to three people with knowledge of the efforts. POLITICO does not state in the excerpt which lawmakers or offices were approached, whether any of them have responded, or what else the draft contains beyond the threshold for blocking a model's release or use and the three risk categories it names.