Anthropic draft proposal would let courts order shutdowns of high-risk AI models
In brief
- Anthropic sent U.S. lawmakers a draft proposal on suspending high-risk AI models, per Crypto Briefing.
- Listed thresholds include over $1 billion in damage, or 50 or more deaths or serious injuries.
- Commerce Department could seek federal court orders to shut down models over specific threats.
- The proposal hasn't been adopted, and Congress hasn't acted on it.
What the draft proposes
The thresholds are specific. According to Crypto Briefing's report, the draft listed risk thresholds that included financial damage of more than $1 billion, or 50 or more deaths or serious injuries. The report presents those as alternatives ("or"), not as a combined test.
So who decides when a model crosses the line? Under the draft as described, the Commerce Department would make the case and ask a federal court for an order. Those orders would apply to specific threats (the report gives biological weapons and cybersecurity breaches as examples), not to AI models in general.
It's still only a draft.
The proposal hasn't been adopted, and Congress hadn't acted on it as of publication. Crypto Briefing didn't cite any statement from Anthropic explaining why it put the proposal forward, and LeoDex News couldn't review the document itself, so the details above rest on that single report.
A prediction market moved (with caveats)
Crypto Briefing also pointed to a prediction market asking which company will have the best AI model by the end of October 2026. The report doesn't name the platform hosting it. In that market, Anthropic's probability fell to 21.5% from 24% over 24 hours, per the report.
The publication read the drop as a sign that market participants were worried about the regulatory environment. That's Crypto Briefing's interpretation, not something the odds show on their own. The report doesn't establish that the draft proposal caused the move, and a 2.5-point swing in a single unnamed market over one day isn't much to build a thesis on.
What comes next
Congress decides what happens next. Crypto Briefing noted that observers are watching whether lawmakers move forward with Anthropic's proposed legislation, and nothing in the report says they've scheduled any action. Until they do, the $1 billion and 50-casualty thresholds (and the court-order mechanism behind them) are proposals, not rules anyone has to follow.
Frequently asked questions
What risk thresholds are in Anthropic's draft AI proposal?
According to Crypto Briefing, the draft listed thresholds that included financial damage of more than $1 billion, or 50 or more deaths or serious injuries. LeoDex News hasn't seen the draft and couldn't independently verify these figures.
How would AI models be shut down under the draft proposal?
The draft would let the Commerce Department ask federal courts for orders to shut down models. Those orders would apply to specific threats, such as biological weapons or cybersecurity breaches, according to Crypto Briefing's report.
Is Anthropic's AI shutdown proposal law?
No. It's a draft proposal sent to U.S. lawmakers, and it hasn't been adopted. As of publication, Congress hadn't acted on it.
