Anthropic has been shopping a draft bill on frontier AI safety around Capitol Hill, according to reporting that began with Politico on Friday 16 October 2026. Reuters, relaying Politico, said Anthropic representatives "quietly pitched" lawmakers on a draft bill to regulate frontier AI safety and security, citing three people with knowledge of the efforts. What follows rests on Reuters, TechCrunch and Tech Insider, which describe Politico's report, along with Anthropic's own June 2026 framework.

The reporting describes a draft rather than introduced legislation. Anthropic's spokesperson confirmed the draft to Politico, TechCrunch reported.

Map showing where Capitol Hill is
Capitol Hill, Denver, United States

What the draft would allow

According to the Reuters account of Politico's report, the government could block a model's release or use only if the model poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. The risk would have to stem from one of three things:

  • the creation or release of a biological weapon;
  • the model evading its developer's controls;
  • a cyberattack.

TechCrunch noted that the 50-people and billion-dollar figures are similar to those in California's Senate Bill 53, which it said was recently passed into law and uses them to define "catastrophic risk". Anthropic's June 2026 framework recommends a definition of catastrophic risk "in line with" the SB 53 definition. That is a recommendation in the framework, not a description of the draft's text.

The evaluators and the court route

Reuters, citing Politico's report, described the mechanism as follows:

  • Independent evaluators would register with the Commerce Department and periodically evaluate a lab and its frontier models.
  • Labs could pay the evaluators, provided the payment is not tied to the outcome of the evaluation.
  • If evaluators determine there is an imminent risk in one of the specified categories for a company's frontier models, they must report it to Commerce within 72 hours of completing their assessment.
  • Commerce could ask a judge to halt a model after determining risk, and a federal judge for the District Court for the District of Columbia could order the suspension.
  • The language explicitly prohibits any other US government department, agency or officer from suspending development, deployment or internal use.

Reuters said the Department of Commerce did not immediately respond to its request for comment.

On the payment arrangement, TechCrunch wrote that it is "not clear how this proposed system would ensure that payments couldn't influence the evaluation". Anthropic's framework, published in June 2026, says an independent evaluator should not have a financial interest in the developer. It also proposes government or pooled funding so that evaluators can work while remaining financially independent of any given developer. The reports do not say how the draft squares the two.

Anthropic's account

An Anthropic spokesperson, confirming the draft to Politico, said, according to TechCrunch: "This bill represents the proposals outlined in Anthropic's Advanced AI Framework, which we have been advocating for with federal and state lawmakers since June." Reuters reported the spokesperson as saying the bill would give the government the "ability to block or deter dangerous deployments of the most powerful models". The spokesperson also said it would require developers to test frontier AI models, be transparent about their findings and have them independently evaluated.

The framework proposes obligations for developers with models trained on more than 10²⁵ FLOP that either earn more than $500 million a year in AI-derived revenue or spend more than $1 billion a year on AI research and development. It names four risk categories: biological weapons, offensive cyber operations, loss of control, and automated research and development. The reports of the draft bill name three.

Where it stood, and the reaction

Tech Insider reported on 17 October that the draft had not been formally introduced in Congress and that Anthropic had pitched it privately to Capitol Hill staff. On 19 October it said coverage had shifted from what the bill contains to whether the $1 billion and 50-person bar is set in the right place. It said Mother Jones had framed that bar as alarming in a 17 October report.

TechCrunch, on 20 October, set the proposal against the wider industry. It wrote that the risks in the draft are "a little bit far afield" from the incidents that have been top of mind for the tech industry, where it said the primary concern has largely been agentic hacking. It also said Anthropic's suggestion appears to go slightly further than other regulatory proposals from the industry, including Nvidia, whose CEO Jensen Huang, it wrote, recently argued the AI industry did not need any new laws.

What to watch

My view is that the reporting so far describes a pitch rather than a bill. The reported mechanism is narrow, a judge's order after a Commerce request. The reported thresholds are high, and their figures are similar to those that already define catastrophic risk in a California statute. Tech Insider reported that the full bill text had not been publicly released as of 19 October, so the reports are the only guide to how those pieces fit together.

Sources

Anthropic (2026) 'Anthropic's Advanced AI Framework', Anthropic, June 2026, available at: https://www-cdn.anthropic.com/files/4zrzovbb/website/0a58d567024a8b448ff15158ebc3625328dfcc1f.pdf (Accessed: 21 October 2026).

Chen, M. (2026) 'Anthropic's $1B AI Safety Bar Draws Backlash [2026]', Tech Insider, 19 October 2026, available at: https://tech-insider.org/anthropic-ai-safety-bill-threshold-backlash-2026/ (Accessed: 21 October 2026).

Politico (2026) Report on Anthropic's draft AI safety bill, Politico, 16 October 2026, available at: https://www.politico.com/news/2026/10/16/anthropic-ai-safety-01124654 (Cited as reported by Reuters and TechCrunch.)

Reuters (2026) 'Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports', The Hindu BusinessLine, 17 October 2026, available at: https://www.thehindubusinessline.com/info-tech/anthropic-quietly-pitches-its-own-ai-safety-bill-to-lawmakers-politico-reports/article71595297.ece (Accessed: 21 October 2026).

TechCrunch (2026) 'Anthropic pitches AI safety bill that may block a model if 50+ people could be killed', TechCrunch, 20 October 2026, available at: https://techcrunch.com/2026/10/20/anthropic-pitches-ai-safety-bill-that-may-block-a-model-if-50-people-could-be-killed/ (Accessed: 21 October 2026).

Virtanen, E. (2026) 'Anthropic Pitches a $1B AI Safety Bill to Lawmakers [2026]', Tech Insider, 17 October 2026, available at: https://tech-insider.org/anthropic-ai-safety-bill-lawmakers-2026/ (Accessed: 21 October 2026).

Picture credits