Anthropic's proposed AI safety bill sets high bar for US intervention

Legislation would trigger only at $1 billion in damages or 50 deaths, centralizing blocking authority in the Commerce Department.

By LineZotpaper

Published18 October 2026

Read Time2 min

Anthropic is lobbying for federal legislation that would allow the US government to block an AI model's release only if it risks causing over $1 billion in damages or killing or injuring more than 50 people. The proposed bill, details of which were reported by Politico, represents the company's latest effort to shape AI safety regulation around catastrophic, rather than widespread, risks.

Under the proposal, an independent third party registered with the Department of Commerce would periodically evaluate AI labs and their models. The lab may pay the evaluators, provided the payment does not affect their independence. The evaluator would have 72 hours to report a potential covered risk, such as the creation of a bioweapon, an AI-powered cyberattack, or a loss of control over an AI agent, to the department. The department would then present evidence to a federal judge in the District of Columbia to determine whether the release should be halted. Only the Commerce Department would hold this authority.

"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," an Anthropic spokesperson said. "These are sensible policies we believe Congress should act on."

The high thresholds reflect the company's stated focus on extreme risks. Anthropic CEO Dario Amodei has warned of a "swarm" of AI-driven botnets taking over the internet and the company has published research on self-replicating AI worms. The bill arrives alongside other Congressional activity on AI safety, including a proposal to enforce kill switches on powerful models and a separate bill from Senator Bernie Sanders that would impose a 20-year prison term for developing AI superintelligence.

Analysis

Why This Matters

The bill proposes a high threshold for government intervention in AI development, prioritizing disaster prevention over addressing smaller-scale harms.
It centralizes the authority to block AI models in a single federal department and a specific DC court, streamlining the decision-making process.
The legislation represents a concrete clash between industry-developed safety frameworks and more aggressive Congressional proposals.

Background

Anthropic was founded with a focus on AI safety. The Advanced AI Framework, published in June, formed the basis for this bill. It arrives alongside other efforts in Washington, including Congressional proposals for kill switches and criminal penalties for superintelligence development.

Key Perspectives

Anthropic: Argues the bill provides a "sensible" balance, enabling beneficial AI development while establishing a clear triage mechanism for the most dangerous scenarios. Congressional Hardliners (Sen. Sanders): Represents a competing view, with proposals treating superintelligence development itself as an existential crime worthy of severe criminal penalties. Industry Peers (Amodei, Altman, Musk): While agreeing on the need for a "controlled slowdown" or pause, their specific policy prescriptions vary, and the reception of this bill among other labs will be a key test of industry unity on regulation.

What to Watch

Whether the bill attracts bipartisan co-sponsors in the House or Senate.
The reaction from safety advocacy groups to the independent evaluator model.
If a major AI incident occurs below the threshold, potentially creating pressure to lower the bar.
The progress of Senator Sanders' competing superintelligence bill.

Sources

Anthropic’s proposed AI safety bill only kicks in if 50 deaths or $1 billion in damages are on the line · Latest from Tom's Hardware

Checked

12 claims in this story checked against 1 source: 5 supported.

Show each claim

Partly supportedAnthropic’s proposed bill sets a high bar for US intervention.Latest from Tom's Hardware: “Anthropic’s proposed AI safety bill only kicks in if 50 deaths or $1 billion in damages are on the line.”
Partly supportedThe legislation would trigger only at $1 billion in damages or 50 deaths.Latest from Tom's Hardware: “Anthropic is quietly lobbying for an AI safety bill that will allow the government to block an AI model’s release only if it poses a risk of causing $1 billion in damages or killing or injuring more than 50 people.”
Partly supportedThe government could block a model only if it risks over $1 billion in damages or killing or injuring more than 50 people.Latest from Tom's Hardware: “Anthropic is quietly lobbying for an AI safety bill that will allow the government to block an AI model’s release only if it poses a risk of causing $1 billion in damages or killing or injuring more than 50 people.”
SupportedPolitico reported details of the proposed bill.Latest from Tom's Hardware: “According to Politico , an independent third party registered with the Department of Commerce would periodically evaluate the lab and its AI models.”
Partly supportedThe proposal reflects Anthropic’s stated focus on catastrophic rather than widespread risks.Latest from Tom's Hardware: “The company more recently said a self-replicating AI worm could spread without data centers , which is why it listed “existential risks to humanity” as one of its risk factors in its IPO prospectus.”
SupportedAn Anthropic spokesperson said, “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” and “These are sensible policies we believe Congress should act on.”Latest from Tom's Hardware: ““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,” an Anthropic spokesperson told the publication. “These are sensible policies we believe Congress should act on.””
SupportedDario Amodei warned of a “swarm” of AI-driven botnets taking over the internet.Latest from Tom's Hardware: “Anthropic CEO Dario Amodei has warned about a “swarm” of AI-driven botnets taking over the internet in the next 6-to-12 months”
Partly supportedAn evaluator would have 72 hours to report a potential bioweapon, AI cyberattack or loss-of-control risk.Latest from Tom's Hardware: “If there is a chance that any of these events are about to happen, it requires the evaluator to report it to the department within 72 hours”
SupportedThe Commerce Department would present evidence to a federal judge in the District of Columbia, who would decide whether to halt the model’s release.Latest from Tom's Hardware: “the department would have to bring all the evidence in front of a federal judge in the District Court for the District of Columbia, who would then determine if the model release should be halted or not.”
SupportedOnly the Commerce Department would have authority to order a lab to stop releasing a model.Latest from Tom's Hardware: “no other department, agency, or officer in the U.S. can order an AI lab to stop releasing its model, except the Commerce Department.”
Partly supportedAnthropic has published research on self-replicating AI worms.Latest from Tom's Hardware: “The company more recently said a self-replicating AI worm could spread without data centers”
Partly supportedSenator Bernie Sanders has a separate bill that would impose a 20-year prison term for developing AI superintelligence.Latest from Tom's Hardware: “Senator Bernie Sanders (I-Vt.) has even gone beyond that, proposing a 20-year prison term for anyone developing AI superintelligence”