Anthropic Is Pitching Its Own Frontier AI Safety Bill to US Lawmakers, Politico Reports

Anthropic representatives have quietly pitched US lawmakers on a draft bill to regulate frontier AI safety and security, Politico reported on October 16, 2026, citing three people with knowledge of the effort. The draft would let the government block a model's release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages or of killing or seriously injuring more than 50 people.

Politico says it obtained the confidential draft, which it calls a notable effort by a frontier lab to shape the rules for its own technology while Congress debates AI policy and safety. An Anthropic spokesperson told Politico the bill reflects the company's Advanced AI Framework, published in June 2026, which it has promoted to federal and state lawmakers since then.

What is in Anthropic's draft bill?

The draft would require frontier labs, the companies building the most capable general purpose models, to test those models, publish safety frameworks and transparency reports, protect whistleblowers and report critical safety incidents to the Commerce Department, according to Politico.

The power to block a model is narrow. The risk has to come from a biological weapon, a model evading its developer's controls, or a cyberattack, and must cross the $1 billion or 50 person threshold (Politico). Politico notes that the language mirrors California's SB 53, and that none of the AI agent incidents labs disclosed this summer appear to have reached those levels.

Independent evaluators registered with Commerce would periodically review each lab and its frontier models. Labs could pay them as long as payment does not depend on the results, and an evaluator that finds an "imminent enumerated risk" must tell Commerce within 72 hours of finishing its assessment, Politico reported.

Who could stop a model, and how?

Only a federal judge could order a model suspended under the draft, on a request from the Commerce Department, according to Politico. The judge would sit on the US District Court for the District of Columbia.

The text explicitly bars any other US department, agency or officer from suspending a model's development, deployment or internal use (Politico). Politico notes that over the summer, sweeping export controls shut down two of Anthropic's most advanced models after a quick series of calls between administration officials and company executives.

Anthropic floated the court route in June. Its framework lists "court enforcement" as one guard against overreach, with an agency able to seek remedies only through a lawsuit. Going by Politico's description, the draft picks that option and names a single agency and a single court.

How does the draft compare with Anthropic's June framework?

Anthropic says the draft carries out its June framework, but as Politico describes it, the draft is narrower on risk scope and differs on evaluator funding. The table sets the two side by side.

Evaluator funding is the clearest gap. Anthropic wrote on September 18, 2026 that long term funding for evaluators should come from pooled or government sources, and on October 15 that evaluators should not be financially dependent on the company they review. The draft, as reported, lets labs pay directly, as long as payment does not hinge on the outcome. Automated AI research, one of the framework's four risk areas, is also missing from the three block triggers in Politico's account.

Where does it fit in Congress's AI safety debate?

Anthropic's draft joins several federal AI safety proposals, Politico reports. In the House, Reps. Lori Trahan (D-Mass.) and Jay Obernolte (R-Calif.) lead the FRONTIER Act, which would also let the government stop risky deployments and preempt some state laws.

In the Senate, Majority Leader John Thune and Sens. Amy Klobuchar (D-Minn.) and Ted Cruz (R-Texas) have been negotiating a bill on catastrophic risks. Sen. Maria Cantwell (D-Wash.), the top Democrat on the Commerce Committee, and some safety groups want a larger government role in reviewing models, while Rep. Greg Casar (D-Texas) and Sen. Bernie Sanders (I-Vt.) have called for a pause in advanced AI development.

Anthropic has not publicly backed a specific bill in Congress, Politico notes, though it supported state laws including two in California. A day before the report, Anthropic said it had given six researchers from the evaluator METR employee-like access, and noted it had signed a White House accord in which companies pledged to work with independent auditors or evaluators.

What we don't know yet

• Which lawmakers or offices received the draft; neither Politico nor Bloomberg names them.
• Whether any member of Congress plans to introduce it, alone or as part of the FRONTIER Act or the Senate talks.
• Which developers the draft would cover, since Politico's report gives no compute or revenue thresholds.
• How the Commerce Department views the role the draft gives it; it did not immediately respond to Reuters.
• How OpenAI, Google and other labs respond to a bill written by a rival.