Anthropic may draft AI safety bill that could block models if over 50 people could die
Anthropic has drafted a bill, Politico reported, that would let the government block an AI model’s release when a “foreseeable, imminent risk” could cause over 50 people to die or over a billion dollars in property damage. The bill says the catastrophic risk must stem from the creation or release of a biological weapon, the model evading its developer’s controls, or a cyberattack.
Key points
- Anthropic’s draft bill would block a model if over 50 people could die or over a billion dollars in damage.
- Independent evaluators, registered with the U.S. Commerce Department, would have 72 hours to notify a judge to halt a release.
- Anthropic says the proposal follows its Advanced AI Framework, advocated since June; Nvidia’s CEO has opposed new AI laws.
The proposal calls for “independent evaluators” registered with the U.S. Commerce Department to assess a model’s disaster potential, possibly paid by the labs so long as payment is not tied to the evaluation’s outcome. If evaluators determine an imminent catastrophic risk, they would have 72 hours to contact the Commerce Department, which must then request a D.C. judge to halt the model’s distribution; no federal agency could act alone. Anthropic confirmed the draft to Politico, saying it reflects its Advanced AI Framework, advocated since June. The article notes DeepMind’s Demis Hassabis has suggested a FINRA‑style regulator, while Nvidia CEO Jensen Huang recently argued the industry does not need new laws.
The story so far
2 episodes →- Anthropic may draft AI safety bill that could block models if over 50 people could diethis story
Anthropic pitches AI safety bill that may block a model if 50+ people could be killed
TechCrunch AI · 20 October 2026
Loading the full article…
This text was published by TechCrunch AI and written by Lucas Ropek. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Solcis launches Compliance Studio and Policy Engine for AI agents on NVIDIA OpenShell · 1 src
- OpenAI employee arrested for hidden recorder; Google releases Gemini 4 Argon with December price hike · 2 src
- Meta’s new terms may allow it to train AI on news sites for free · 16 src
- IronScales report finds AI agents are new phishing targets · 2 src
- Opinion: Solnit urges Americans to oppose tech oligarchs amid AI backlash · 1 src
Comments
via GitHub Discussions