US lawmakers propose emergency model shutdown and independent AI security audits

AI safetyRegulationUnited States

Published: 23 July 2026. Bipartisan House members proposed an “AI Kill Switch Act” that would let federal authorities order a model shutdown in a defined loss-of-control scenario. A separate six-member bipartisan proposal would require leading model developers to submit powerful systems to Commerce-accredited independent security audits. The White House’s top technology adviser was briefed on, and is monitoring, the OpenAI–Hugging Face incident.

Why it matters: These bills would move containment, audit and emergency response from company policy into public law. They are legislative proposals and may change or fail. Impact: Frontier developers and enterprises should make shutdown, revocation, evidence preservation and external audit technically feasible now. a report · Underlying OpenAI disclosure