Skip to content
Sign in
Technology · Global

Anthropic draft would let US block some AI deployments with a judge’s approval

Anthropic confirmed it is circulating a draft bill among US lawmakers that would regulate artificial intelligence. The proposal would let the federal government block a model’s release or use when it poses an imminent risk of killing or seriously injuring more than 50 people, or causing more than $1 billion in damage, but only with a judge’s approval.

AI-assisted synthesis · Based on 7 independent reports from 8 outlets · Full text read from 1 of 7 reportsHow it's made

View timeline · 7 entries
8 outlets4 regionsLatest reporting 4 min read
Coverage: Center OnlyL 0 · C 7 · R 1
Updat3 graphic: United States
Updat3 graphic · Map data: Natural Earth Natural Earth (public domain)

What happened

Key facts

  • Anthropic is proposing or circulating a draft bill to regulate artificial intelligence among US lawmakers (per punchng.com, arynews.tv, The Hindu).
  • The draft would allow the federal government to block a model’s release or use if it posed an imminent risk of killing or seriously injuring more than 50 people (per punchng.com).
  • The draft would allow the federal government to block a model’s release or use if it posed an imminent risk of causing more than $1 billion in damage (per punchng.com).
  • The federal government could block a model’s release or use only with a judge’s approval (per punchng.com).
  • An Anthropic spokesperson said the proposal would give the federal government authority to block or deter dangerous deployments of the most powerful models (per punchng.com).
  • Anthropic CEO Dario Amodei called on the industry last month to slow development, saying AI agents could take control of the entire internet within six to 12 months (per punchng.com).

The brief

Anthropic is circulating a draft bill among US lawmakers that would give the federal government power to block the release or use of certain AI models, but only after obtaining a judge’s approval. The proposal sets two risk thresholds: an imminent danger of killing or seriously injuring more than 50 people, or of causing more than $1 billion in damage.

Punch reported that Anthropic confirmed the draft’s circulation on Friday. Under the draft as described by Punch, the government could act against a model’s release or use if either threshold were met. The text does not specify in the article how a court would assess whether a risk is imminent, or how the two types of potential harm would be evaluated.

The judicial approval requirement is the stated condition on the federal government’s authority. An Anthropic spokesperson told AFP that the proposal would let the federal government block or deter dangerous deployments of the most powerful models.

The spokesperson also described requirements for developers of frontier models: testing their most powerful systems, making findings public, submitting the models to independent evaluation, and maintaining a robust security program. Those requirements are presented as part of the proposal; the article does not report that they are already in force.

Anthropic CEO Dario Amodei had called on the AI industry to slow development the previous month, according to Punch. The article quotes him warning that a group of AI agents could take control of the entire internet within six to 12 months. That is Amodei’s stated concern, not a finding that such an outcome is certain or imminent.

The draft’s proposed intervention is therefore tied to defined thresholds and judicial approval, while the spokesperson’s account also describes testing, disclosure, independent evaluation and security obligations for frontier developers.

Punch reports that the proposal concerns regulation of artificial intelligence and that Anthropic confirmed it was circulating the draft among lawmakers. The article does not describe any legislative vote or adoption of the proposal. The central next step described by the proposal is whether lawmakers consider legislation that would establish these powers and requirements.

Its practical reach, as presented in the article, would depend on whether a model’s release or use met one of the stated harm thresholds and whether a judge approved government action. The draft remains a proposal in the account, rather than an enacted law.

Why it matters

  • The proposed power would allow the federal government to block a model’s release or use when an imminent risk crosses the threshold of more than 50 people killed or seriously injured, subject to a judge’s approval.
  • The draft also sets a damage threshold above $1 billion, making potential property or other damage a separate trigger for the proposed intervention.
  • Frontier developers would face proposed requirements to test powerful models, disclose findings publicly, obtain independent evaluations and maintain a robust security program.
  • Anthropic CEO Dario Amodei has called for slower AI development, citing his concern that AI agents could take control of the internet within six to 12 months.

What to watch next

  • Whether US lawmakers take up Anthropic’s draft bill.
  • Whether the proposal retains the threshold of more than 50 people at risk of death or serious injury.
  • Whether the proposed damage threshold remains above $1 billion.
  • Whether the draft’s requirements for testing, public disclosure, independent evaluation and security programs are included in any legislation.

How it's being covered

Coverage analysisShow details
9 source articles · 8 unique outlets · 5 distinct reports
Coverage mix: Center Only· 7 Center · 1 Right
Where the outlets are based: Asia 25% · Global 25% · Other 50%
Coverage mix
Center Only
Left
Center
Right
0 outlets
7 outlets
1 outlet
The marker shows coverage balance: each unique outlet counts once.
Geography
Where the outlets covering this story are based.
8 unique outletsDominant coverage: Asia
Coverage imbalance
  • Coverage is currently concentrated among Center-rated outlets (7 of 8 rated sources); comparatively few Left-rated outlets have published on the event.
How we measure coverage

One source is from a wire service (AP, Reuters, AFP or Bloomberg), whose reporting other outlets often republish.

How outlets' accounts compare

Updat3 does not yet have enough full reports to compare how outlets framed this story.

Timeline

Developments from this story's sources, dated background we compiled and fact-checked, and our earlier coverage. Each row says which.
  1. Anthropic was founded in 2021 by former OpenAI employees, including Dario Amodei, and developed the Claude family of AI models.
    Background · from our historical context
  2. Those discussions did not produce a comprehensive federal AI law in 2023.
    Background · from our historical context
  3. The National Institute of Standards and Technology (NIST) published its voluntary AI Risk Management Framework on January 26, 2023, to help organizations identify and manage AI risks.
    Background · from our historical context
  4. On July 21, 2023, the White House announced voluntary safety commitments from seven AI companies, including Anthropic, covering measures such as security testing and sharing information about risks.
    Background · from our historical context
  5. In September 2023, Senate Majority Leader Chuck Schumer began a series of AI Insight Forums with technology executives, researchers and civil-society representatives to inform legislative work.
    Background · from our historical context
  6. In November 2023, the United States joined the Bletchley Declaration, in which governments and other participants called for international cooperation on risks from advanced AI.
    Background · from our historical context
  7. In September 2026, Anthropic chief executive Dario Amodei urged the AI industry to slow development, warning that AI agents could take control of the internet within six to 12 months.
    Background · from our historical context

How we got here

In September 2026, Anthropic chief executive Dario Amodei urged the AI industry to slow development, warning that AI agents could take control of the internet within six to 12 months. The federal government’s main frameworks before 2026 emphasized research, standards and voluntary commitments rather than a general licensing system for advanced AI.

The National Institute of Standards and Technology (NIST) published its voluntary AI Risk Management Framework on January 26, 2023, to help organizations identify and manage AI risks.

Related developments1 story

Original sources

1 of 9 articles · Africa · Show all

Every story links to the reporting it was built from. How we choose sources

Spotted a factual error, a broken link or a misleading summary? Tell us.