Anthropic floats bill letting judges halt risky AI models
The draft would let the government halt a rogue model only when 50 or more people face imminent serious harm.
By Standard Topic. AI-generated from the reporting listed under Coverage. Published . Updated .
This story is no longer being updated. It reflects reporting up to .
What happened
Anthropic is circulating a draft U.S. AI safety bill that would let courts block AI models posing imminent catastrophic risk, using a threshold of 50 deaths or $1 billion in damages. The proposal places the final decision with a judge after a formal review process. Anthropic has now pitched the draft to lawmakers, and critics say it is too weak.
Why it matters
The threshold would determine when the government could shut down a model. One report says a model with a foreseeable $900 million risk could still fall short of the bar for federal intervention.
Timeline
- Oct 20, 8:00 PM ET: Anthropic's draft bill was reported to block AI models only if they risk 50 deaths or $1 billion in damages, with a judge's approval required.
- Oct 20, 11:46 PM ET: The draft was described as letting U.S. courts stop model releases after independent assessments identify an imminent threat of mass casualties or major property damage.
- Oct 21, 11:07 AM ET: Anthropic pitched the draft to lawmakers, and critics said the bill was too weak.
- Oct 21, 1:43 PM ET: Analysis of the draft found that a model posing a foreseeable $900 million risk could still miss the bar for federal intervention.
Coverage
- Anthropic backs bill that would allow 50 deaths before AI model shutdown (thecooldown.com)
- Anthropic Writes Its Own Rules for Shutting Down AI (PYMNTS)
- Anthropic Wants More Government Oversight on Its Own A.I. Models (cheddar.com)
- Anthropic bill would pause AI models if 50+ people could die, report says (mashable.com)
- Rogue AI Must Threaten 50 Lives Before Being Shut Down, Anthropic Suggests (novaramedia.com)
- Anthropic's AI Safety Bill Hides Judge-Only Block Clause (The Tech Buzz)
- Anthropic pitches AI safety bill that may block a model if 50+ people could be killed (techcrunch.com)
- Anthropic’s Proposed Regulations for AI Include a Death Threshold of at Least 50 People Before Halting Models (truthout.org)
- Anthropic’s leaked AI bill sparks backlash (Yahoo News)
- Anthropic's draft AI bill would restrict shutdowns to extreme cases
- Anthropic pitches lawmakers AI safety bill with model-blocking powers
- Anthropic proposes federal AI law letting Washington block risky models
- Anthropic pitches Congress an AI safety bill as co-founder urges rules
More from Technology & AI
- China puts humanoid robots to work serving food and sorting parcels
- Newsom proposes sweeping AI safety rules for Washington
- Google releases Gemini 4 Neon AI model with million-token context
- Microsoft and Intel widen their cloud chip partnership
- Google DeepMind rolls out Genie 4 to select subscribers
- AI helped AMD shorten chip design work, OpenAI says