Anthropic proposes legislation to halt AI if it poses a risk to 50 people
Anthropic proposes a legislative framework to halt AI models if they pose a risk to 50 people or cause damages worth $1 billion.
Translated by Qiyas from الجريدة — Original source
Anthropic has proposed a new legislative framework to regulate AI safety in the United States, which includes the possibility of halting the release of advanced models if they pose an imminent risk that could lead to the death or serious injury of more than 50 people, or cause property damage exceeding $1 billion.
According to Politico, the draft bill specifies circumstances under which intervention can be taken to prevent the deployment or use of AI models if the risk is linked to the creation of biological weapons, the launch of cyberattacks, or if the models exceed the safety controls established by their developers.
The draft proposes subjecting AI companies to periodic assessments conducted by independent evaluators registered with the U.S. Department of Commerce, while allowing companies to fund these assessments, provided that the financial compensation is not tied to the results.
Under the proposal, evaluators must notify the U.S. Department of Commerce within 72 hours of identifying an imminent catastrophic risk. The department can then seek a court order to halt the release or use of the model, while the draft prevents other federal agencies from unilaterally deciding to stop it without judicial recourse.
The proposed standards resemble provisions in California legislation concerning catastrophic risks arising from AI.
A spokesperson for Anthropic confirmed that the project reflects the proposals outlined in the company’s advanced AI safety framework, noting that it has urged lawmakers at both the federal and state levels to adopt these measures since June last year.
