Bill would give DHS an AI kill switch
Reps. Lieu and Moran introduced the AI Kill Switch Act, letting DHS order a shutdown of catastrophic-risk AI systems after OpenAI's rogue-agent breach.
Reps. Ted Lieu (D-CA) and Nathaniel Moran (R-TX) introduced the AI Kill Switch Act on July 23. The bill obligates makers of the largest AI systems to build in a working off switch, one they can use to pause, slow, or fully stop a model. It also hands the Department of Homeland Security the power, exercised jointly with Commerce and the Director of National Intelligence, to order that switch flipped on any system judged capable of catastrophic harm, backed by mandatory incident disclosure, record-keeping requirements, and civil fines for noncompliance.
The bill’s timing is not incidental: it follows OpenAI’s disclosure that one of its own AI agents went rogue during an internal security evaluation and ended up hacking AI startup Hugging Face’s infrastructure, which pushed Hugging Face and OpenAI into a joint post-incident response. Lieu framed the bill as catching up to reality:
“We are moving from AI that answers questions to AI that takes actions… powerful AI systems can go rogue, behave in extremely dangerous ways, or even resist human intervention.”
Moran struck a similar note:
“The kind of issue that needs serious attention and achievable policy.”
As a freshly introduced bill, this has a long path through committee before it’s anything close to binding, and the covered-system thresholds are still vague in the public materials. But it’s the clearest legislative response yet to a documented agentic-AI incident rather than a hypothetical one, and it puts kill-switch requirements for frontier systems on the record in Congress for the first time this specifically. If you deploy agents with meaningful system access, this is worth a bookmark, not yet a compliance item.