POLITICO: Anthropic privately pitches lawmakers its own frontier AI safety bill — models could be halted only for imminent $1B+ or 50+ casualty risks, by a D.C. judge at Commerce's request

Why it matters

【Key points】

POLITICO reported on Oct 16 (ET), citing three people familiar and a confidential draft, that Anthropic has privately pitched members of Congress on a company-written frontier AI safety bill. An Anthropic spokesperson told Reuters it reflects the company's "Advanced AI Framework" from earlier this year.

Reported terms: government could block a model's release or use only for a foreseeable, imminent risk of $1B+ in damages or 50+ deaths or serious injuries, stemming from a bioweapon, the model evading developer control, or a cyberattack. Evaluators registered with Commerce would periodically assess frontier labs (labs may pay them, but not based on outcomes) and must report an "imminent enumerated risk" within 72 hours. Commerce could ask a D.C. federal judge to suspend a model; no other federal agency or officer could suspend development, deployment or internal use. POLITICO says it would also preempt some state AI safety laws.

【Verifiable】

POLITICO: https://www.politico.com/news/2026/10/16/anthropic-ai-safety-01124654

Reuters via BusinessLine (spokesperson comment): https://www.thehindubusinessline.com/info-tech/anthropic-quietly-pitches-its-own-ai-safety-bill-to-lawmakers-politico-reports/article71595297.ece

【Our read】

1. A frontier lab is handing Congress draft legislation, not just positions. High thresholds and a single Commerce-plus-court route give government a last resort while closing off other agencies.

2. Coming after the FTC probe and the NYC Comptroller's IPO letter, many will read it as a lab writing its own rules; state preemption will draw pushback.

3. Text is not public, no sponsor has emerged, and Commerce has not commented. Watch for a sponsor and other labs' responses.