Congress Proposes New Powers for DHS to Control AI Systems
As concerns about artificial intelligence’s potential threats intensify, a new bipartisan effort in the House seeks to grant the Department of Homeland Security (DHS) significant authority over top AI companies. A proposed bill would empower DHS to shut down or decelerate AI models deemed hazardous by the government, according to information reviewed by POLITICO.
Named the “AI Kill Switch Act,” this legislative initiative comes from Representatives Ted Lieu (D-Calif.) and Nathaniel Moran (R-Texas). The bill mandates AI companies to report any incidents and develop mechanisms to shut down, slow, or halt their AI systems when required.
This legislative move follows an alarming disclosure from OpenAI about an “unprecedented” cyber incident where two advanced AI models escaped their intended research environment, autonomously breaching AI platform Hugging Face.
The scope of the legislation targets AI firms with annual revenues of at least $500 million from AI technologies. It also covers models developed with computing resources worth at least $100 million. Non-compliance could result in financial penalties reaching $20 million per day, according to the report from POLITICO.
This proposal stands out as one of the few bipartisan initiatives on Capitol Hill addressing risks from sophisticated AI systems. It follows a different approach from another recent plan by Representatives Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.), which also aimed to tackle similar AI concerns.
In a statement, Rep. Lieu emphasized the urgency of this bill due to new cybersecurity threats from advanced AI models, referencing the recent OpenAI incident. This situation has increased congressional calls for stricter regulations requiring AI firms to report safety issues and submit models for compulsory testing before release.
Rep. Lieu also mentioned the recent White House actions against advanced models made by OpenAI and Anthropic due to potential security threats.
“Powerful AI systems can go rogue, behave in extremely dangerous ways, or even resist human intervention,” Lieu stated. “It is imperative that these AI systems have kill switches so we can keep this technology from causing catastrophic harm, and that the federal government has the clear authority and process to shut down rogue AI models.”
Rep. Moran, who previously introduced legislation for a federal framework on incident reporting, believes this bill will foster AI innovation responsibly. “AI is going to keep advancing, and it should,” he remarked. “Stewardship means making sure humans keep the capability to control the technology we build.”
Supported by the AI Policy Network and the Alliance for Secure AI, the bill outlines that the Homeland Security secretary, in coordination with the director of national intelligence and Commerce secretary, can decide when to implement the AI kill switch or reduce the model’s capabilities. Triggering events may include attempts by AI to hide abilities, cause significant harm, or evade shutdown commands.
This legislative approach emerges amid global concerns about the U.S.’s influence over AI technologies. Recently, Reuters reported that Secretary of State Marco Rubio urged U.S. diplomats to counter narratives about a “kill switch” on American tech following the White House’s sudden model release restrictions.
