US lawmakers are preparing legislation that would require AI companies to shut down or throttle systems when ordered by the Department of Homeland Security, according to The Verge’s report on a Politico story. Representatives Ted Lieu and Nathaniel Moran are expected to introduce the AI Kill Switch Act.
The proposal points to a growing concern in Washington: if advanced AI systems create urgent safety or security risks, the government wants a direct way to slow or stop them. That idea is simple in phrase but complicated in practice.
A kill switch could mean different things depending on whether a system is accessed through a hosted API, embedded in consumer products, deployed inside a company, or released as open weights. The bill will also have to define who decides a system is dangerous, what evidence is required, and how companies can challenge or comply with an order.
The debate is likely to expose a divide between emergency preparedness and operational reality. Regulators want leverage over high-risk systems, while developers will want clear standards that do not turn routine incidents into broad shutdown powers.