Anthropic lobbies for narrow federal powers to halt advanced AI systems

Anthropic quietly handed Capitol Hill lawmakers a draft bill that would let the government block powerful AI models only in cases of extreme harm. Reported by Politico on 16 October 2026, the text sets the bar at foreseeable risk of death or serious injury to more than 50 people or damage exceeding $1 billion.

Only the Department of Commerce could seek a court order to stop development or deployment. No other agency would have authority to intervene. Covered threats include bioweapons, loss of control, and major cyberattacks.

In exchange, developers would have to run catastrophic-risk tests, release transparency reports, and use independent evaluators. The rules would apply to models trained above 10²⁵ FLOP when a company earns over $500 million yearly from AI or spends more than $1 billion on research.

The proposal builds on Anthropic’s June 2026 Advanced AI Framework. The firm spent $3.13 million on federal lobbying in 2025 and donated $20 million to an advocacy group pushing stricter AI oversight.

Based on information from cryptonews.net