Anthropic backs AI safety bill with high bar for blocking model releases
The proposal would shift any stop order to a narrow federal process, leaving only extreme risks eligible for intervention.
Anthropic is backing a draft AI safety bill that would let the U.S. government stop a model’s launch only under extreme harm thresholds, including at least $1 billion in damage or more than 50 deaths or injuries. The proposal also covers risks from a bioweapon, AI-assisted cyberattacks, or a lab losing control of a model or agent. An independent evaluator registered with the Commerce Department would periodically assess labs and models, and if it saw an imminent qualifying risk it would have to report within 72 hours. After that, the Commerce Department would be required to take the evidence to a federal judge in the District of Columbia, and no other U.S. agency or official could order a release stopped. Anthropic says the bill reflects the company’s Advanced AI Framework and that it has been pushing similar ideas with federal and state lawmakers since June.
Why it matters
If adopted, the bill would make release stops rare and tightly controlled. That means model makers would face a single federal channel for urgent action, rather than a wider set of officials able to intervene. It also shows Anthropic is trying to turn its own safety framework into public policy.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.
Sources
- Tom's Hardware