Anthropic wants Congress to set strict limits on when federal agencies can prevent advanced AI models from reaching users. A confidential draft law received from POLITICALLY would condition government intervention on specific catastrophic threats and require court approval for suspensions.
The proposal would provide pioneering AI developers with a clearer legal framework while limiting the circumstances that could trigger government action. It also raises questions about how much oversight regulators should exercise over increasingly powerful systems.
Strict thresholds for intervention
Under the proposal, federal authorities could seek to stop a model only if it poses a foreseeable, imminent threat. Qualifying risks must involve damages of more than $1 billion or the death or serious injury of more than 50 people.
The bill limits these threats to three scenarios. These include the development or release of biological weapons, models that escape the control of their developers, or cyberattacks. Other forms of harm would not fall within this specific intervention framework.
Anthropic says the legislation reflects the Advanced AI Framework, which has been promoted to federal and state lawmakers since June. “These are common-sense guidelines that we believe Congress should act on,” an Anthropic spokesperson reportedly said.
Independent assessors would monitor compliance. They would register with the Department of Commerce and regularly evaluate pioneering AI developers and their models. Reviewers could receive compensation from the companies they audit, provided their compensation does not depend on their results.
If an assessment identifies an imminent threat covered by the bill, assessors must notify Commerce within 72 hours of completing their review.
Courts would control suspensions
The proposal would prevent the Commerce Department from shutting down a model on its own. Instead, the department would have to seek an order from a federal judge in the District Court of the District of Columbia.
The draft also prohibits other federal ministries, agencies and officials from suspending the development, deployment or internal use of a model. This restriction could reduce the risk of sudden executive action against AI developers.
The issue is particularly relevant after US export controls reportedly disabled two advanced Anthropic models over the summer. The incident sparked concerns across the tech industry about government officials restricting access to powerful systems.
However, the bill would also preempt certain state requirements related to public risk disclosure, independent assessments, and critical incident reporting. That provision could reignite disputes over whether Washington should set a national standard or maintain stronger federal protections.
Congress is weighing competing proposals
Anthropic’s draft sparks a heated legislative debate over how the United States should regulate border AI. Reps. Lori Trahan and Jay Obernolte have advanced the FRONTIER Act, which would authorize government intervention against risky models and override some state laws.
In the Senate, Majority Leader John Thune has been working with Senators Amy Klobuchar and Ted Cruz on legislation to address catastrophic AI risks. But some safety advocates and Sen. Maria Cantwell want a more rigorous federal review process.
Other lawmakers have pushed for broader restrictions. Representative Greg Casar and Senator Bernie Sanders have called for halting the development of advanced AI and banning superintelligence.
Anthropic’s proposal takes a narrower approach. Border developers would have to create security frameworks, publish transparency reports for new models, protect whistleblowers and report critical incidents to the trade.
Whether lawmakers accept these limits will help determine how much authority the federal government gains over advanced AI systems. The debate also focuses on the question of whether a single national framework can protect the public without giving regulators excessive discretion.
