Full Breakdown
Anthropic Pushes Draft Bill to Regulate Frontier AI Safety
By Drooid · · How we work
Anthropic’s Draft AI Safety Bill Presented to Congress
On October 16, representatives of the artificial-intelligence company Anthropic met privately with Capitol Hill lawmakers to share a confidential draft bill aimed at governing “frontier” AI models. The effort, reported by multiple outlets, marks the first known instance of an AI developer formally proposing legislation that would give the federal government authority to halt the release or use of a model deemed dangerously risky.
Core Provisions of the Proposal
The draft specifies two quantitative thresholds for government intervention: a foreseeable, imminent risk of more than $1 billion in damages or a risk of killing or seriously injuring more than 50 people. Eligible risks must arise from the creation or release of a biological weapon, the model evading its developer’s controls, or a cyber-attack.
To enforce these standards, the bill would create a regime of independent evaluators registered with the Department of Commerce. Evaluators would periodically assess frontier models and could be paid by the labs, provided compensation is not tied to the outcome of the assessment. If an evaluator determines an “imminent enumerated risk,” they must notify the Commerce Department within 72 hours. The department could then petition a federal judge in the District of Columbia to order a suspension of the model. The legislation explicitly bars any other U.S. agency from independently halting development, deployment, or internal use.
Anthropic’s Official Position and Government Response
An Anthropic spokesperson described the draft as a translation of the company’s “Advanced AI Framework,” which the firm has been advocating to state and federal officials since June. The spokesperson called the measures “sensible policies” that Congress should adopt. The Department of Commerce has not yet commented on the proposal.
Potential Impact on AI Regulation
If enacted, the bill would give the federal government a direct tool to block the most powerful AI systems based on narrowly defined risk criteria, a model that mirrors provisions in California’s recent AI legislation (SB 53). By mandating independent testing and rapid reporting, the proposal could set a precedent for how frontier AI is overseen nationally, potentially shaping future legislative approaches to emerging technologies.