Oct 16, 2026 — Anthropic representatives pitched Capitol Hill lawmakers on draft federal frontier-AI safety legislation, per three people familiar with the effort (Politico).
Intervention thresholds: Government may block model release/use only for foreseeable, imminent risk of >$1B damages or death/serious injury to >50 people from bioweapon creation/release, cyberattacks, or models evading developer controls.
Mechanism: Commerce-registered independent evaluators assess labs periodically; labs may pay evaluators if payment is not tied to outcomes. Imminent enumerated risk must be reported to Commerce within 72 hours. Commerce may seek D.C. federal court order to suspend a model. Language would bar other federal agencies from suspending development/deployment/internal use.
Anthropic said the bill reflects its Advanced AI Framework advocated since June 2026. Draft legislation — not enacted law.