Anthropic is shopping a federal bill on frontier AI safety that would let the government step in only when a model risks over $1 billion in damages or the death or serious injury of more than 50 people.
Anthropic is pitching federal lawmakers a bill it wrote itself to regulate frontier AI. Under the draft, the government could block a model's release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people.
The AI incidents this summer, including ones involving OpenAI's and Anthropic's own software, would fall below that threshold and be exempt from intervention, even though some of those acts are felonies.
Writing your own liability ceiling into law is a remarkably convenient kind of regulation.
Curated from high-quality sources, with concise summaries and key takeaways.
