# Anthropic pitches a $1bn threshold for blocking AI models

> Anthropic has reportedly pitched lawmakers a bill allowing AI model blocks only above $1bn damage or 50 casualties.

*A draft bill shown to US lawmakers would let the government seek to halt a model only for extreme harms, and only through a federal judge.*

By Bez · InsidersFeed
Canonical: https://insidersfeed.com/news/anthropic-pitches-a-1bn-threshold-for-blocking-ai-models

Anthropic is trying to set the terms on which the US government could stop an AI model from reaching the market, and it has drawn the line high. The company has privately presented lawmakers with a draft bill that would allow a block only when a model threatens more than $1 billion in damage or death or serious injury to more than 50 people, according to [TechSpot](https://www.techspot.com/news/114252-anthropic-wants-ai-models-blocked-only-if-they.html). The report says the story originated with Politico. The draft is a proposal, not law.

The trigger is a foreseeable, imminent risk above those thresholds. Only three kinds of risk would qualify: biological weapons, models escaping their developers' control, and cyberattacks. Harms outside those categories would not meet the test, however large.

The process is as notable as the threshold. The Commerce Department would have to make its case in federal court, and a judge in the US District Court for DC would decide whether to stop a release. One report says the draft would expressly prohibit any federal agency from halting distribution of a model on its own. Another account says it would limit state interventions to the most severe scenarios.

An Anthropic spokesperson described the package as "sensible policies we believe Congress should act on."

Critics argue the judicial gate makes blocking so procedurally difficult that agencies might never try. The details come from secondary summaries of the draft. The bill's text and its sponsors have not been seen.

## The Insider view

The pitch is notable for who it protects. A narrow trigger, three named risk types and a court hurdle leave nearly every commercial release outside the reach of a federal stop order. Frontier labs, Anthropic included, would gain predictability: a model could be pulled only after the government wins before a judge.

The exposed party is the regulator. Commerce would carry the burden of proof, with no power to act alone, and critics say that may deter it from acting at all. State governments also stand to lose room to intervene outside the gravest scenarios.

Watch for who sponsors the bill, since none is yet known, and whether the published text matches these summaries. Until then, the proposal shows where a safety-focused lab wants the line drawn, and that line is high.

## Key takeaways

- Anthropic's draft sets a high bar: over $1bn in damage or more than 50 casualties.
- Commerce would have to win in federal court, and agencies could not halt releases alone.
- Critics say the court gate may deter regulators from ever trying to block a model.

## Sources

- [Anthropic wants AI models blocked only if they threaten over $1 billion in damage or more than 50 deaths or serious injuries](https://www.techspot.com/news/114252-anthropic-wants-ai-models-blocked-only-if-they.html) — TechSpot, 2026-10-21