Anthropic proposes a draft bill to block dangerous AI models, requiring a high threshold for intervention and third-party evaluations.
Anthropic proposes AI safety legislation
Anthropic, a major AI company, has submitted a draft bill to U.S. lawmakers that would allow the government to block AI models considered dangerous, including some of the most advanced ones currently available.Conditions for blocking AI models
The draft bill specifies that the government can only block an AI model if it poses a predictable and imminent risk of over one billion dollars in damages or serious harm to more than 50 people. The risk must arise from one of three specific causes: the development of a biological weapon, an AI model escaping its developer’s control, or a cyberattack.Legal process for blocking models
Under the proposal, the Department of Commerce would have the authority to request a federal judge in the District of Columbia to suspend a model once a risk is identified. The bill also prevents other federal agencies or officials from interfering with the development or use of AI models.Requirements for AI developers
The draft bill mandates that AI developers conduct rigorous testing of their models, share their findings transparently, and have their models evaluated by independent third-party experts. These evaluators would be registered with the Department of Commerce and would inspect laboratories and models regularly.Legislative and political context
Anthropic’s proposal draws from its earlier Advanced AI Framework, which it has promoted to lawmakers since June. However, the White House has opposed mandatory AI regulations, and recent reports suggest that any such legislation is unlikely to become law soon.
