What happened
Anthropic published 'Anthropic's Advanced AI Framework' — a position paper laying out what governments should do in the near term about the most serious risks from frontier AI. It is the basis for a draft frontier AI safety-and-security bill Anthropic pitched to members of Congress. The draft would let the government block a model's release/use only on a 'foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people' arising from a biological weapon, the model evading its developer's controls, or a cyberattack; would task independent evaluators (registered with Commerce) to assess frontier labs and report 'imminent enumerated risk' within 72 hours; would bar other US agencies from suspending model development; would preempt some state AI laws; and would require labs to publish safety frameworks, transparency reports, whistleblower protections and critical-incident reports to Commerce. The framework and draft bill were publicly described in the Nov 16, 2026 Faegre Drinker regulatory briefing and discussed across the week as Anthropic prepared its record IPO.
Why it matters
This is a frontier lab's concrete draft national AI safety statute — it effectively sketches the template for future federal AI legislation and signals where industry is willing to draw trigger thresholds, evaluator oversight, and preemption lines.
Action needed
Brief the policy/regulatory and risk teams on the draft bill's enumerated-risk triggers and evaluator-reporting requirements, and map them against your own AI risk taxonomy and incident-reporting obligations.