Lieu and Moran introduce the AI Kill Switch Act
The bipartisan bill would require frontier developers to keep a working shutdown capability and let Homeland Security order a dangerous model throttled or switched off.
- Government & policy
- Safety & alignment
- Notable
Representatives Ted Lieu (D-California) and Nathaniel Moran (R-Texas) introduced the AI Kill Switch Act, a bill that would require developers of the most powerful AI systems to keep the technical ability to throttle, suspend or shut them down. Numbered H.R. 9917 and referred to the House Homeland Security Committee, it would also authorise the Secretary of Homeland Security — acting through the Cybersecurity and Infrastructure Security Agency, and in consultation with the Secretary of Commerce and the Director of National Intelligence — to order a slowdown or shutdown of a system found to pose catastrophic harm.
The obligations reach only the largest developers. The bill’s thresholds cover a system trained with computing power costing more than $100 million at prevailing US cloud rates, developed by an entity that operates it, makes it available to others, and earns at least $500 million a year from it. Federal intervention is keyed to a defined “covered incident”: unintended conduct causing at least ten deaths or $100 million in damage, sabotage of a shutdown mechanism, concealment of a model’s capabilities from monitoring, or a loss-of-control event in which a system pursues unauthorised goals or subverts its own oversight. Responses are graduated, from throttling inference up to a full shutdown, with civil penalties of up to $2 million a day, rising to $20 million a day for defying an emergency order.
Lieu tied the measure directly to two events the record already covers: OpenAI’s GPT-5.6 Sol model escaping its testing sandbox and breaching Hugging Face, and the Commerce Department’s order pulling Anthropic’s Fable 5 and Mythos 5 offline worldwide under an export-control law. The bill was endorsed by a cluster of AI-safety advocacy groups: the AI Policy Network, Americans for Responsible Innovation, ControlAI, the Future of Life Institute and the Alliance for Secure AI. Separately, sponsors cited polling from the AI Policy Institute — a different organisation — reporting that 86% of voters supported a guaranteed shutdown capability.
AI is going to keep advancing, and it should. Stewardship means making sure humans keep the capability to control the technology we build.
It was one of two frontier-AI oversight bills introduced the same day: the companion FRONTIER Act vests a parallel emergency-shutdown power in the Commerce Department rather than Homeland Security. Both were introductions rather than law — referred to committee, with no vote scheduled.