Anthropic proposes AI safety bill with high damage thresholds

Anthropic has put forward a proposed federal AI safety bill that would set a high bar for regulatory intervention, triggering oversight only in cases involving 50 or more deaths or $1 billion or more in damages, according to a report from Tom’s Hardware. The outlet, citing the proposal’s structure, states that the framework would route enforcement primarily through the U.S. Commerce Department, effectively positioning it as the main agency with authority to halt a deployed AI model under the bill’s terms.

The report frames the thresholds as central to the proposal’s design, noting that incidents falling below these casualty and damage figures would not activate the bill’s intervention mechanisms. Tom’s Hardware’s coverage does not specify additional enforcement tiers or lower-threshold reporting requirements within the proposal as described.

The piece was surfaced via Hacker News on October 18, 2026, and credits the original reporting to Tom’s Hardware for details of the bill’s structure and its concentration of regulatory authority within Commerce. No additional government agencies are named in the material as holding comparable intervention power under the proposal.

The Specola notes the figures and the agency structure as reported, without further elaboration beyond what the cited coverage establishes. The proposal’s current status, legislative sponsorship, and timeline were not detailed in the available source material.