Story perspectives

Anthropic Drafts Bill to Block Deadly AI Models

By Drooid · Oct 22, 2026, 6:25 PM UTC · How we work

Summary

Story summary

Anthropic drafted a U.S. bill that would let the government block AI models causing over 50 deaths or $1 billion damage.
The bill routes enforcement through the Department of Commerce, which must obtain a District of Columbia federal judge’s order.
Independent organizations registered with Commerce would evaluate models and report imminent catastrophic risks within 72 hours.
Anthropic’s federal affairs head Anthony Cimino said the thresholds mirror California law and would not shield companies from liability.
AI safety experts Nathan Calvin, Nat Purser and Nate Soares warned the proposal would weaken safeguards and limit federal oversight.

History

Historical Context

Congress responded to rising cyber threats by passing the Computer Security Act in 1987, assigning NIST and the Department of Defense to set federal security standards. Earlier, the 1986 Computer Fraud and Abuse Act criminalized unauthorized computer access. Anthropic’s current proposal echoes that framework, asking the Department of Commerce to review AI models before release, potentially extending pre-emptive federal oversight.

1 of 2

Story summary

Anthropic drafted a U.S. bill that would let the government block AI models causing over 50 deaths or $1 billion damage.
The bill routes enforcement through the Department of Commerce, which must obtain a District of Columbia federal judge’s order.
Independent organizations registered with Commerce would evaluate models and report imminent catastrophic risks within 72 hours.
Anthropic’s federal affairs head Anthony Cimino said the thresholds mirror California law and would not shield companies from liability.
AI safety experts Nathan Calvin, Nat Purser and Nate Soares warned the proposal would weaken safeguards and limit federal oversight.

1 / 2