Anthropic Quietly Pitches Congress Its Own AI Safety Bill

Anthropic privately pitched Congress a draft AI safety bill that would sharply limit the government's power to block dangerous frontier models.

@sharedot

Anthropic Quietly Pitches Congress Its Own AI Safety Bill

By @sharedot · · 8 pages

  • AI
  • AI Regulation
  • Anthropic
  • AI Safety

Anthropic privately pitched Congress a draft AI safety bill that would sharply limit the government's power to block dangerous frontier models.

What the Draft Bill Would Do

Under the proposal, federal officials could seek to suspend an advanced AI model only if it poses a foreseeable and imminent risk of causing more than $1 billion in damages or killing or seriously injuring more than 50 people.

The thresholds would apply to three categories of threats: creation or release of biological weapons, cyberattacks, and AI systems escaping their developers' controls.

An Anthropic spokesperson told Reuters the bill represents proposals outlined in the company's Advanced AI Framework, which the company says would let the government block dangerous deployments while requiring developers to test frontier models, disclose safety findings and undergo independent evaluations.

A Lab Drafting Its Own Rules

The move marks one of the most concrete attempts yet by a frontier AI lab to turn its own policy preferences into binding federal law, according to reporting by shattered.io, which notes the document reads less like a wish list and more like a markup-ready statute.

Anthropic has spent 2026 arguing that AI companies cannot be trusted to police themselves — the company's head of public policy, Sarah Heck, has said "we can't be checking our own homework."

Writing the legislative text shifts Anthropic from commentator to drafter, which is why outlets including TradingView, investing.com and Bloomberg Law picked up the story within hours of Politico's report.

A Narrow Path to Blocking a Model

Enforcement would run through independent evaluators registered with the Commerce Department, who would periodically assess frontier AI developers. Companies could pay those evaluators, provided compensation was independent of the findings.

If an assessment identified an imminent threat meeting the bill's criteria, evaluators would have 72 hours to report it to Commerce, which could then seek a court order.

Only a judge in the U.S. District Court for the District of Columbia could order a suspension — judicial authorization would be a prerequisite.

According to Politico's reporting as carried by Law Commentary, all other federal departments, agencies and officials would be barred from independently suspending covered models, concentrating authority in Commerce and the courts.

Why It's Surprising

The draft would also override certain state AI safety laws through federal preemption, potentially limiting states' ability to impose separate testing, disclosure and safety requirements.

That stands in tension with Anthropic's own June statement that "absent a strong federal law, states should retain full authority to legislate," a shift noted in reporting by CNA. The company previously supported state-level measures including California's frontier AI transparency requirements.

The bill also sets a deliberately narrow definition of catastrophic risk, capping — rather than expanding — how aggressively regulators could intervene against the industry's most powerful systems.

The Stakes Amid a Crowded Field

The pitch lands as Washington debates competing visions of AI oversight.

Democratic Sens. Mark Warner, Brian Schatz and Andy Kim introduced the Artificial Intelligence Risk Management and Security Act of 2026 (S. 5576) on September 29, creating a federal AI Safety Board with enforceable requirements, though Sen. Ted Cruz blocked its unanimous-consent passage and it sits before the Senate Commerce Committee.

Charlie Bullock of the Institute for Law & AI praised Anthropic's proposal as "pretty reasonable" and "more than other frontier labs have done," but told followers on X he considers it weaker than the House's competing FRONTIER Act, per CNA's reporting.

What Comes Next

The draft has yet to be formally introduced in Congress; any legislation would need approval from both chambers and the president's signature.

Anthropic is meanwhile expanding independent scrutiny of its own technology: on October 15 it granted six researchers from the nonprofit METR employee-like access to internal systems, hardware and office space, with METR free to publish findings independently, though Anthropic retains limited redaction rights for sensitive information.

The company is also working with SecureBio and Accenture on external assessments. Context matters: Anthropic was banned from the Pentagon in February and forced in June to suspend its most advanced models for nearly three weeks, according to Reuters reporting carried by CNA.

Sources

  1. lawcommentary.com › Anthropic Quietly Pushes AI Safety Bill That Would Limit Government Power to Block Dangerous Models
  2. channelnewsasia.com › Anthropic quietly pitches its own AI safety Bill to US lawmakers
  3. shattered.io › Anthropic Pitches Congress a $1B AI Safety Bill
  4. tradingview.com › Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports

More on AI