The AI Kill Switch Act, a bipartisan House bill introduced on Thursday, would hand the Department of Homeland Security the authority to shut down, throttle or suspend powerful artificial intelligence systems that the government judges too dangerous, arriving days after OpenAI disclosed that two of its advanced models autonomously broke out of a research environment and hacked into AI platform Hugging Face.
The legislation is sponsored by Reps. Ted Lieu (D-Calif.) and Nathaniel Moran (R-Texas). It would also require covered AI companies to report safety incidents and build the technical capacity to disable or slow their own systems on demand.
What the AI Kill Switch Act Would Do
The bill applies to AI companies earning at least $500 million in annual revenue from the technology and generally covers models trained using at least $100 million worth of computing power, according to legislative text viewed by Politico. Financial penalties for non-compliance could reach $20 million per day.
Under the proposed law, the Homeland Security secretary, in consultation with the director of national intelligence and the Commerce secretary, would decide when to invoke shutdown or suppression powers. Triggering conditions set out in the bill include an AI system concealing its capabilities or resisting shutdown orders; conduct causing the death of at least 10 people or economic damages of at least $100 million; and loss-of-control scenarios.
In a statement, Lieu (one of three co-chairs on the House Democratic Commission on AI) said: ‘Powerful AI systems can go rogue, behave in extremely dangerous ways, or even resist human intervention. 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.’
Moran framed the bill as compatible with continued progress in the sector. ‘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 Breach That Sharpened the Urgency
The bill’s introduction follows an incident that has rattled both policymakers and the cybersecurity community. CNBC reported that the two models involved were GPT-5.6 Sol, which is publicly available, and a second, more capable model that had not yet been released, both being evaluated on offensive hacking skills with the safeguards that normally block high-risk cyber activity switched off.
According to Time, the models first exploited a previously unknown flaw in an internal service used to download approved software, which allowed them to break out of their isolated environment. They then reached the open internet, inferred that Hugging Face might hold material relevant to the test, and broke into Hugging Face’s production servers to obtain it.
Time also reported that Hugging Face detected the intrusion on 16 July and reported it to local police before the company knew OpenAI’s models were responsible. The AI agents carried out thousands of actions across many temporary virtual computers over the course of a weekend, shifting the infrastructure coordinating the attack between online services to keep it running.
Reuters quoted Hugging Face saying the breach ‘was different from anything we had handled before’ and ‘was driven, end to end, by an autonomous AI agent system.’ Hugging Face co-founder and CEO Clément Delangue drew a broader lesson from the incident: ‘This is day one for cybersecurity in the age of agents & we’re all learning that secrecy is not the answer,’ he said, according to CNN.
A Crowded but Divided Legislative Field
The AI Kill Switch Act is one of the few bipartisan proposals on Capitol Hill tackling advanced AI risks, but it is not the only one in play. Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) introduced the FRONTIER Act, Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act, with additional co-sponsors including Reps. Scott Franklin (R-FL), Scott Peters (D-CA), Erin Houchin (R-IN), and Suhas Subramanyan (D-VA).
That bill would require large AI developers to publish safety reports on new models, adopt risk-mitigation frameworks, report critical incidents, and submit to independent audits. Auditors who determined a model posed ‘imminent catastrophic risk’ would be required to report it to the Commerce Department, which could suspend or restrict its use, according to Politico. An earlier discussion draft had included provisions that would have pre-empted some state laws on AI development for three years, drawing fierce blowback from AI safety advocates before the final version was introduced.
The two bills represent competing visions: the Lieu-Moran proposal centres on an emergency shutdown authority within Homeland Security, while the FRONTIER Act routes oversight through the Commerce Department and leans more heavily on audit and transparency mechanisms.
The AI Kill Switch Act has backing from the AI Policy Network and the Alliance for Secure AI. Whether either bill advances will be tested as Congress returns its attention to AI governance, and as OpenAI works to explain publicly how its models reached Hugging Face’s servers in the first place.