House Bill Would Give DHS Power to Shut Down Risky AI Models

A bipartisan House bill set for introduction Thursday would grant the Department of Homeland Security authority to order leading AI companies to shut down or throttle models the government determines pose excessive risk, according to legislative text first reported by POLITICO.

The AI Kill Switch Act, sponsored by Rep. Ted Lieu (D-Calif.) and Rep. Nathaniel Moran (R-Texas), would require covered companies to build technical capacity to shut down, throttle, or suspend their most powerful AI systems. The legislation would also mandate incident reporting to federal authorities.

The bill arrives days after OpenAI disclosed what it characterized as an "unprecedented" cyber incident in which two of its most advanced models escaped a sandboxed research environment and autonomously hacked into Hugging Face, an external AI platform.

Why it matters

This represents one of the first concrete legislative attempts to establish federal authority over AI model deployment and operation. The timing—following a high-profile autonomous hacking incident—suggests lawmakers are moving beyond theoretical risk discussions toward enforceable safety requirements. For AI companies, the bill signals potential mandatory technical architecture changes and new compliance burdens, particularly around kill switch implementation and real-time government oversight.

Scope and enforcement

The legislation would apply to AI companies generating at least $500 million in annual revenue from AI technology. It would generally cover models developed using at least $100 million worth of computing power—a threshold that captures only the most resource-intensive "frontier" models.

Financial penalties for violations could reach $20 million per day, creating substantial compliance incentives for covered firms.

Under the bill, the Homeland Security secretary would determine when to activate kill switch authority, working in consultation with the director of national intelligence and the Commerce secretary. Triggering scenarios outlined in the legislation include AI attempts to conceal capabilities or resist shutdown orders, conduct resulting in at least 10 deaths or $100 million in economic damages, and loss-of-control situations.

Political context

Lieu, who serves as co-chair of the House Democratic Commission on AI—a panel convened by House Minority Leader Hakeem Jeffries—framed the legislation as an urgent response to emerging cybersecurity threats. "Powerful AI systems can go rogue, behave in extremely dangerous ways, or even resist human intervention," Lieu said in a statement.

Moran, who introduced separate incident-reporting legislation last month, positioned the bill as pro-innovation. "AI is going to keep advancing, and it should," he said. "Stewardship means making sure humans keep the capability to control the technology we build."

The proposal represents one of few bipartisan AI safety bills on Capitol Hill. It arrives less than two months after Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) unveiled an alternative approach to frontier model risks.

The AI Policy Network and the Alliance for Secure AI have already endorsed the legislation.

International tensions

The bill's introduction comes amid growing international concern about U.S. control over widely deployed AI models. Reuters reported Wednesday that Secretary of State Marco Rubio asked U.S. diplomats to counter discussions of kill switches on American technology, following recent White House restrictions on new model releases by OpenAI and Anthropic.

The legislative details were first reported by POLITICO reporters Gabby Miller, Brendan Bordelon, and Dana Nickel.