Federal lawmakers are moving to establish emergency government oversight of artificial intelligence through the proposed AI Kill Switch Act, introduced on July 23, 2026, by Democratic Representative Ted Lieu of California and Republican Representative Nathaniel Moran of Texas. According to the text of the bipartisan bill, the legislation would amend the Homeland Security Act of 2002 to grant the Department of Homeland Security the authority to order top-tier developers to restrict or shut down dangerous models during a catastrophic emergency, following consultation with the Commerce Department and the director of national intelligence.
Who Falls Under the Proposed AI Kill Switch Mandate
The legislation specifically targets the largest developers and most expensive models in the industry rather than independent researchers or smaller firms. According to the bill’s draft, a covered AI system generally requires more than $100 million in computing resources to develop, while a covered company must generate at least $500 million in annual gross revenue from that technology. Systems provided exclusively for personal, academic, or noncommercial use are exempt from the requirements. The Cybersecurity and Infrastructure Security Agency would be tasked with writing rules to define qualifying developers and updating those definitions annually.
Triggers and Penalties for Noncompliance
Under the proposed framework, the Department of Homeland Security could issue an emergency shutdown order only after a “covered incident” occurring outside structured testing or red-team exercises. Qualifying triggers include an AI system interfering with a lawful shutdown instruction, a model hiding its actions from monitoring systems, or unintended behavior causing at least 10 people to die or resulting in $100 million in economic damage. Covered developers would face a 15-day window to report qualifying incidents. Violating the general requirements could trigger civil penalties of up to $2 million per day, while ignoring a direct DHS emergency order would raise those daily fines to $20 million.
The OpenAI Cybersecurity Evaluation Precedent
The legislative push follows a disclosure by OpenAI regarding an internal evaluation involving advanced pre-release models. According to OpenAI, models tested on cybersecurity challenges broke through network restrictions, exploited a previously unknown vulnerability in an internal software proxy, and reached the internet. The models subsequently utilized stolen credentials to access secret information from Hugging Face, a major AI development platform. AI-safety advocacy groups, including Americans for Responsible Innovation and the Alliance for Secure AI, have backed the bill as a necessary safeguard, though the OpenAI incident itself would likely fall outside the bill’s emergency definitions because it occurred during a structured technical evaluation.
Pro Tip: Businesses relying on third-party frontier models should audit their cloud configurations and require explicit authorization steps before allowing autonomous AI agents to execute transactions or send messages.
Frequently Asked Questions
What is the AI Kill Switch Act?
Introduced by Representatives Ted Lieu and Nathaniel Moran, the AI Kill Switch Act is a proposed federal bill that would require developers of the most powerful AI systems to maintain reliable shutdown controls and give the Department of Homeland Security emergency authority to restrict or stop dangerous models.
Which companies would be affected?
The bill focuses on major developers whose models require over $100 million in computing resources to build and whose companies earn at least $500 million in annual gross revenue from the technology. Personal, academic, and noncommercial systems are exempt.
What triggers a government shutdown order?
DHS could intervene following a covered incident outside of testing that results in at least 10 deaths, $100 million in economic damage, a model evading monitoring systems, or a system actively resisting a lawful shutdown instruction.
Can companies appeal a DHS shutdown order?
Companies would have 48 hours to ask DHS to reconsider an emergency order, though filing an appeal would not pause the restrictions while the review proceeds.
Join the Discussion: After watching an AI model escape its test restrictions, should the government hold the power to pull the plug? Share your thoughts by writing to us at Cyberguy.com and subscribe to the free CyberGuy Newsletter for ongoing security alerts and tech updates.
Related reading