Anthropic Quietly Pitched Lawmakers a Draft AI Safety Bill, Politico Reports

Anthropic has quietly pitched lawmakers on Capitol Hill a draft bill to regulate the safety and security of frontier AI, Politico reports, citing three people with knowledge of the effort. The draft would let the US government block a model’s release or use only in narrow, severe cases, and only through a federal judge in Washington. It is a proposal, not a filed bill. No lawmaker has been reported as introducing it.

What the draft reportedly says

The details come from Politico’s report, as summarised by Reuters. TSN has not seen the draft text.

According to the report, the government could 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”. That risk would also have to come from one of three sources: the creation or release of a biological weapon, a model evading its developer’s controls, or a cyberattack.

The draft reportedly sets out an oversight process:

• Independent evaluators would register with the Commerce Department and periodically evaluate AI labs and their frontier models. Labs could pay the evaluators, provided payment is not tied to the outcome of the evaluation.
• A 72-hour reporting rule. If evaluators find an “imminent enumerated risk” in a company’s frontier models, they must report it to Commerce within 72 hours of completing their assessment.
• Court-only suspension. Commerce could ask a judge to halt a model after determining a risk, but only a federal judge of the US District Court for the District of Columbia could order a suspension.
• Other agencies barred. The language explicitly prohibits any other US government department, agency or officer from suspending development, deployment or internal use of a model.

What Anthropic says

An Anthropic spokesperson told Reuters the bill reflects proposals in the company’s “Advanced AI Framework”, released earlier this year. It would give the government the “ability to block or deter dangerous deployments of the most powerful models”. The spokesperson said it would also require developers to test frontier models, be transparent about their findings, and have them independently evaluated.

Anthropic’s published framework, dated June 2026 in its full PDF, sets out the same broad ideas. These are obligations on frontier developers to test for catastrophic risks such as biological weapons, cyber operations and loss of control, to engage external evaluators, and to disclose risks and incidents, alongside government power to block dangerous deployments with “concrete safeguards” against misuse of that power.

The Commerce Department did not immediately respond to Reuters’ request for comment.

Why it matters

The draft would give Washington real power to stop a model, but within tight limits. The high damage and casualty thresholds, the restriction to three categories of risk and the requirement for a single federal court all narrow when that power could be used. Barring other agencies from suspending models would also stop regulators from acting outside that process. Supporters may see that as protection against heavy-handed regulation. Critics could see it as a high bar written by a company that would itself be regulated.

It is unusual for an AI developer to draft and circulate its own legislative text rather than just comment on lawmakers’ bills. Anthropic has argued publicly that transparency alone is not enough to address the most serious AI risks. The draft shows what that position could look like in law.

What we don’t know

• Which lawmakers Anthropic pitched, and whether anyone plans to introduce the bill.
• The full text, including how “frontier model” and “imminent enumerated risk” are defined.
• How the Commerce Department or other AI companies view the proposal.

What this means

• Politico reports Anthropic pitched lawmakers a draft AI safety bill. It is a proposal and has not been introduced.
• Models could be blocked only for imminent risks of more than $1 billion in damages or more than 50 deaths or serious injuries, from bioweapons, cyberattacks or loss of control.
• Evaluators would report such risks to Commerce within 72 hours, and only a DC federal judge could suspend a model. Anthropic says the draft reflects its Advanced AI Framework.

Sources

• Politico: Anthropic AI safety bill report
• Reuters via StreetInsider: Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports
• Anthropic: Advanced AI Framework
• Anthropic: Advanced AI Framework (PDF, June 2026)

Related stories

• EU Is Investigating OpenAI and Anthropic After AI Agents Escaped Testing, Euronews Reports; Virkkunen Says Enforcement Has Started
• EU Tech Chief Virkkunen Warns of Under-Tested AI Models and Says Brussels Has Sent Information Requests to OpenAI, Anthropic and Chinese Firms
• AI Rules This Week: Anthropic Rewrites Its Rulebook, and the UK Regulator Turns to AI Agents