Artificial intelligence company Anthropic has reportedly approached US lawmakers with a draft bill aimed at regulating the safety and security of advanced AI systems, including proposals that would allow the government to block models considered an imminent threat.
According to a report published by Politico on Friday, Anthropic representatives privately presented the proposed legislation to lawmakers on Capitol Hill, citing three people familiar with the discussions.
Under the draft bill, the US government could block the release or use of an AI model only if it poses a foreseeable and imminent risk of causing more than $1 billion in damage, or killing or seriously injuring more than 50 people.
The report said such risks must arise from the creation or release of a biological weapon, an AI model bypassing its developer’s safeguards, or a cyberattack.
An Anthropic spokesperson told Reuters that the proposal reflects ideas contained in the company’s Advanced AI Framework, which was released earlier this year.
The spokesperson said the bill would give the government the “ability to block or deter dangerous deployments of the most powerful models.”
The proposed legislation would also require developers of advanced AI systems to test their models, disclose their findings and submit them for independent evaluations.
Under the draft, independent evaluators would register with the US Department of Commerce and periodically assess AI companies and their most advanced models. The evaluators could receive payment from the companies they assess, provided their compensation was not linked to the outcome of the evaluation.
If an evaluator identifies an “imminent enumerated risk” involving a company’s advanced AI models, the evaluator would be required to notify the Commerce Department within 72 hours of completing the assessment.
The proposal would also establish limits on the government’s authority to intervene. According to the report, the Commerce Department could ask a judge to suspend an AI model after determining that it presents a qualifying risk.
However, the draft explicitly prohibits other US government departments, agencies and officials from independently suspending the development, deployment or internal use of such models.
A federal judge at the US District Court for the District of Columbia would have the authority to order a suspension.
The US Department of Commerce did not immediately respond to a request for comment on the reported proposal.
The draft bill reflects growing efforts to establish rules for advanced AI systems as governments and technology companies grapple with potential risks involving cybersecurity, biological threats and the ability of powerful models to operate beyond their developers’ intended controls.
