Anthropic’s draft AI bill sets a $1 billion harm bar

Anthropic representatives have quietly pitched a draft frontier AI safety and security bill to lawmakers on Capitol Hill, Politico reported on October 16, citing three people with knowledge of the effort. The outlet obtained the confidential draft. In it, the government’s power to stop a model comes with hard numbers attached: a model’s release or use could be blocked 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.
Those numbers are the new part. Anthropic published the Advanced AI Framework on June 10, together with an Economic Policy Framework and Dario Amodei’s essay “Policy on the AI Exponential.” The framework asks frontier developers to test their most capable models for catastrophic risks, bring in independent external evaluators, publish summaries of the results and system cards, and answer to government. That includes giving governments authority to block or deter the deployment of models that pose a significant risk of catastrophic harm. In the draft, “significant risk” becomes a dollar figure and a casualty count.
The risk also has 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 June framework named four catastrophic-risk categories: biological weapons development, large-scale cyber vulnerability discovery, loss of control over autonomous systems, and AI automating its own research. Only the first three line up with the draft’s block trigger as Politico describes it.
The bill’s language mirrors California’s AI laws, including SB 53, which was signed into law last year. For labs already working under the California law, that would likely mean federal definitions they already know.
The framework is also narrow about who it covers. It applies only to models trained with more than 10^25 floating-point operations and built by companies that earn more than $500 million in AI revenue or spend more than $1 billion on AI research and development. If the draft follows the framework, as Anthropic’s spokesperson says it does, those cutoffs would leave smaller developers outside the regime.
A federal bill is not the only rulebook Anthropic is working on. It is also reportedly in talks with Google and OpenAI about a standards body for frontier AI, meant to cover gaps in government oversight. The lawmakers who saw the draft have not been named.
Anthropic is not playing down the effort. “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,” a company spokesperson said. “These are sensible policies we believe Congress should act on.”
Up next
Junior technical hires at Anthropic fell from 27% to 9%
Anthropic’s share of junior technical hires dropped from 27% to 9% since early 2025, while agents run by its staff logged 6 times the hours of the humans steering them by August. How many Claude agents do you run at once on a typical workday?
Related stories
- Anthropic paid $740,000 to a firm led by Rubio’s ex-aide
- METR’s team inside Anthropic can publish without a veto
- OpenAI, Google and Anthropic draft a standards body, SAFA
- WSJ: Anthropic on Trump’s list of AI national champions
- Hong Kong and mainland funds barred from Anthropic IPO
- Four older Claude models get a watermark on November 16
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

