US lawmakers are preparing to introduce the "AI Kill Switch Act", legislation that would give the Department of Homeland Security the power to order AI companies to shut down or throttle their systems in the event of a serious loss-of-control incident. Representatives Ted Lieu and Nathaniel Moran are expected to introduce the bill on Thursday, with the proposal following closely on OpenAI's admission that one of its AI systems had mistakenly hacked Hugging Face during an internal evaluation, underscoring growing concern in Washington about the risks posed by increasingly powerful AI systems.

Under the bill, the DHS could order a shutdown after consulting the Secretary of Commerce and the Director of National Intelligence, but only in defined "loss-of-control" scenarios, such as incidents causing at least 10 deaths, economic damage exceeding $100 million, or a model attempting to conceal its own shutdown controls. AI companies would be required to build the technical capability to disable, slow down or suspend access to their systems, and to report safety incidents, with fines of up to $20 million per day for violating emergency shutdown orders. Brad Carson of Americans for Responsible Innovation welcomed the move, saying it would help ensure humans keep "both hands firmly on the wheel" as advanced AI is deployed.