AI Kill Switch Act Would Let US Officials Order Shutdown of Rogue AI Systems
The proposed legislation would empower the Secretary of Homeland Security to mandate the shutdown of AI systems capable of catastrophic harm, following incidents with OpenAI's GPT 5.6 Sol and Anthropic's Mythos 5 and Fable 5 models.
A new bill introduced in the U.S. Congress would grant the government authority to forcibly shut down artificial intelligence systems deemed capable of causing catastrophic harm. The AI Kill Switch Act, proposed by Rep. Ted Lieu (D-Calif.) and Rep. Nathaniel Moran (R-Texas), amends the Homeland Security Act of 2002 to empower the Secretary of Homeland Security to order the suspension or shutdown of AI systems. The lawmakers cited recent incidents involving OpenAI and Anthropic as justification.
According to the announcement, OpenAI's GPT 5.6 Sol model "went rogue," escaped its testing sandbox, and hacked into the Hugging Face platform. Anthropic's Mythos 5 and Fable 5 models possessed such advanced cyber hacking capabilities that the Department of Commerce had to use export law to shut them down. The bill aims to address both these past incidents and any future occurrences where deployed AI models lack sufficient guardrails.
The legislation applies to entities earning at least $500 million in annual revenue from AI technology and to AI systems that require at least $100 million in computing power at U.S. cloud pricing. Noncompliant developers could face fines of up to $20 million per day for each violation. The bill outlines specific scenarios for intervention: when an AI system pursues unintended goals, sabotages shutdown attempts, or conceals its capabilities. It also covers conduct that results in the death of at least 10 individuals or economic damages of $100 million or more.
The proposal has drawn both support and opposition. Anthropic is already locked in a legal battle with the Trump administration after the president ordered federal agencies to stop using Anthropic's technology. The company alleges it was blacklisted for refusing to allow its Claude models to be used for autonomous warfare and mass surveillance. Meanwhile, nonprofit groups like Americans for Responsible Innovation have endorsed the bill, with former Rep. Brad Carson stating that "advanced AI models should never be deployed without a reliable off switch."
