Artificial intelligence company Anthropic has presented a draft bill to lawmakers in the United States, proposing new legal requirements to strengthen the safety and security of advanced AI systems.

The proposal seeks to establish government oversight of powerful AI models and introduce independent assessments to identify risks before the systems are released or deployed.

Under the proposed legislation, the government could block the release or use of an AI model only if it posed a foreseeable and imminent risk of causing more than $1 billion in damage or killing or seriously injuring more than 50 people.

The specified risks would have to arise from the creation or release of a biological weapon, an AI model evading its developer’s controls or a cyberattack.

The draft would also require independent evaluators registered with the US Department of Commerce to assess AI developers and their advanced models periodically.

AI companies would be permitted to pay the evaluators, provided their payments were not linked to the outcome of the assessments.

Where an evaluator identified an imminent risk covered by the proposed legislation, it would be required to notify the Department of Commerce within 72 hours of completing its assessment.

The department could then ask a federal judge to suspend the release or use of the model. The proposal reportedly limits the authority to suspend a model to a federal court, preventing other government departments, agencies or officials from independently halting its development, deployment or internal use.

An Anthropic spokesperson said the draft reflected proposals contained in the company’s Advanced AI Framework, which it had been advocating for with federal and state lawmakers since June.

The spokesperson said the framework would require AI developers to test their models, disclose their findings and allow independent evaluators to review their systems regularly.

The company argued that the proposed rules would give the federal government the ability to prevent dangerous deployments of the most powerful AI models.

The initiative comes amid growing concern among policymakers and technology companies about the potential risks of increasingly capable AI systems, including their possible misuse in cyberattacks and the development of biological weapons.

However, the proposal remains a draft, and its provisions would require legislative consideration before becoming law. The US Department of Commerce had not immediately commented on the reported proposal.