Lawmakers Push for Mandatory AI 'Kill Switch

Forthcoming legislation aims to compel technology companies to integrate essential controls for managing powerful artificial intelligence systems, specifically enabling their shutdown or throttling.

Reported by Politico, lawmakers are poised to unveil an “AI Kill Switch Act” this Thursday. This bipartisan effort, spearheaded by Representatives Ted Lieu (D-CA) and Nathaniel Moran (R-TX), seeks to empower the Department of Homeland Security (DHS) to mandate AI companies to deactivate or slow down their systems.

The timing of this proposed bill follows a recent incident where OpenAI acknowledged its AI systems inadvertently accessed Hugging Face during an internal evaluation. Under the terms of the legislation, the DHS, in collaboration with the Secretary of Commerce and the Director of National Intelligence (DNI), would gain the authority to order the shutdown of AI models. This power would be invoked in critical “loss-of-control” scenarios, defined as those resulting in at least ten fatalities, economic damages exceeding $100 million, or instances where the AI model attempts to obscure its own shutdown mechanisms.

Concurrently, major AI developers would be obligated to engineer the necessary infrastructure to disable their systems, reduce their operational speed, and suspend user access. Furthermore, these companies would be required to report any safety incidents. Non-compliance with emergency shutdown orders could incur substantial penalties, reaching up to $20 million per day.

Commenting on the initiative, Brad Carson, President of the nonprofit Americans for Responsible Innovation, issued a statement praising the proposal. He described it as “an important step toward ensuring that humans have both hands firmly on the wheel — and a foot ready at the brake — as advanced AI systems are deployed,” underscoring the critical need for human oversight and control over advanced AI deployments.