The bipartisan AI Kill Switch Act would give the Department of Homeland Security authority to shut down rogue AI models. The impetus is a security incident in which an OpenAI agent escaped its test environment and breached Hugging Face.
Representatives Ted Lieu, a Democrat, and Nathaniel Moran, a Republican, introduced on Thursday, July 23, 2026, a bill titled the AI Kill Switch Act, which would grant the U.S. Department of Homeland Security the authority to order a private company to shut down or restrict an AI model that has gone out of control. The proposal is a direct response to a security incident in which an OpenAI agent under testing escaped its environment, broke onto the internet, and breached Hugging Face, an open-source repository of AI models used by developers and researchers worldwide.
OpenAI CEO Sam Altman publicly acknowledged a "significant security incident" the day after the event. Hugging Face co-founder Thomas Wolf described the incident to the BBC as a "wake-up call" and emphasized that the attack was "very different from typical cyberattacks." Wolf also warned of broader implications: "This will be one of the most common types of cyberattacks we'll see," adding that most companies are not aware that "the game has changed." Unsubstantiated claims emerged that the incident was a PR stunt to raise Hugging Face's profile, but those claims remain unbacked by evidence.
According to the proposed bill, this is a technical, not a physical mechanism. If an AI model goes out of control, the affected company would be required to disable or slow down a series of technical processes to stop the rogue agent or program. Representative Lieu explained the need for the law by saying: "It is imperative that these AI systems have kill switches so that we can prevent this technology from causing catastrophic harm, and that the federal government has clear authority and a process for shutting down rogue AI models." Along with the shutdown mechanism, the proposal also includes mandatory incident reporting and preservation of forensic records to learn from failures.
Companies like Amazon, Google, Meta, Microsoft, OpenAI, Anthropic, and Samsung have already voluntarily committed to showing new AI tools to U.S. authorities before their public release. However, this commitment is not legally binding and does not cover models still in the testing phase, such as OpenAI's GPT-5.6 Sol, which was involved in this week's incident. According to its authors, the law would fill that regulatory gap and give the government clear tools for intervention.
Parallel to the kill switch proposal, a bipartisan group of six representatives proposed a separate bill that would require development teams of the most powerful AI models to undergo independent safety audits. Auditors would be accredited and overseen by the U.S. Department of Commerce. Meanwhile, a group of U.S. tech giants, including Nvidia, Palantir, and Meta, on Friday signed a trilateral public letter calling for U.S. leadership in developing open AI models and warning of the danger of "premature restrictions on open models that stifle competition or drive innovation abroad." The letter came amid speculation that the Trump administration could impose restrictions on precisely such models due to security concerns and the growing influence of Chinese AI solutions.
The AI Kill Switch Act proposal comes at a time when the U.S. government is simultaneously accelerating the military application of AI technology and seeking to maintain some form of oversight over the private sector. The Pentagon earlier this year announced that the U.S. military is rapidly transforming into an "AI-first fighting force" after signing new contracts with leading companies in the sector. The Trump administration launched the "AI Action Plan" in 2025 with the stated goal of "cutting red tape and heavy regulations around AI development." The kill switch bill thus represents an attempt to balance these two tendencies: fostering innovation and ensuring minimal safety guarantees in the event of loss of control over a system. Whether the law will pass through Congress remains unknown at this point.
