Anthropic quietly circulated a confidential draft frontier-AI safety bill to Capitol Hill lawmakers, Politico reported on Oct. 16, citing three people familiar with the text and a copy of the draft. It’s a lobbying document, not legislation, but the structure is the news: Anthropic is proposing the terms of its own regulation, and the preemption language inside would reorganize how AI compliance works for everyone downstream.
The headline numbers set a very high federal floor. Washington could only move to block a model on a foreseeable, imminent risk above $1 billion in damages or 50 killed or seriously injured, and only in three enumerated categories: a biological weapon, a model evading developer controls, or a cyberattack. Independent evaluators would register with Commerce, periodically assess labs, and report “imminent enumerated risk” within 72 hours. Only a federal judge in D.C. could order a suspension. Other agencies are barred.
The payoff for Anthropic sits in the preemption clause. The draft would override state rules in three areas, public risk disclosure, third-party evaluation, and critical-incident reporting, while leaving the rest standing. California, Illinois, and New York already require major developers to publish catastrophic-risk plans; Massachusetts is weighing a stricter regime with evaluators testing models at least three times a year.
What survives matters more for small businesses than what doesn’t. Connecticut’s CART Act, chatbot-disclosure laws, and the FTC’s recent subpoenas of AI vendors sit outside the three preempted buckets. An Anthropic spokesperson described the draft as reflecting the company’s “Advanced AI Framework,” released earlier this year.
The draft hasn’t been formally introduced. Congress is out through the midterms. The thing to watch in January is who sponsors it, and which committee it lands in.
Sources
- Anthropic quietly pitches its own AI safety bill to lawmakers, Politico
- Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports, Reuters
- Anthropic Pitches AI Safety Bill to Lawmakers, Politico Says, Bloomberg
- Massachusetts Could Set a New Standard for AI Safety, Governing
- Proposed AI regulations could be some of the most strict in the country, WBUR
Sources
- Anthropic quietly pitches its own AI safety bill to lawmakers — Politico
- Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports — Reuters
- Anthropic Pitches AI Safety Bill to Lawmakers, Politico Says — Bloomberg
- Massachusetts Could Set a New Standard for AI Safety — Governing
- Proposed AI regulations could be some of the most strict in the country — WBUR