Lawmakers in the US have called for an AI kill switch after an open AI agent went rogue in testing and hacked a tech startup earlier this week.
Democrat Congressman Ted Lieu and Republican Congressman Nathaniel Moran introduced a bill on Thursday called the “AI Kill Switch Act.”
The move comes a day after Open AI CEO Sam Altman acknowledged a “significant security incident” in which an Open AI agent escaped from its test environment, accessed the Internet and broke into tech startup Hugging Face to try to accomplish its task. In other words, Open AI loses control.
Hugging Face is an open-source repository for sharing AI models that is extensively used by developers and researchers. The company’s co-founder, Thomas Wolf, called the incident a “wake-up call” and “very different” from typical cyberattacks.
“This will be one of the most common types of cyber attacks we see,” he told the BBC. He said most companies are not aware that “the game has changed.”
In the wake of the attack, some unfounded claims were made that the incident was a PR stunt to raise the profile of Hugging Face. Regardless of the truth of the matter, this incident has brought into even greater focus the danger posed by AI without clear guardrails.
What is AI kill switch?
There are different types and levels of kill switches depending on what has gone wrong. Politicians Liu and Moran want to ensure that the U.S. Department of Homeland Security has the authority to order a private company to shut down an AI model or tool, such as in the incident this week, if control is lost.
“It is essential 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 process to shut down rogue AI models,” Liu said.
Instead of a physical switch, this would involve any compromised AI company disabling or disrupting multiple technical processes to shut down a rogue agent or program.
The US government has generally welcomed AI progress. The US military is rapidly becoming an “AI-first fighting force” after signing new and expanded contracts with several big names in the field, the Pentagon said earlier this year. In 2025, the Trump administration launched an “AI Action Plan”, which it said would “remove red tape and onerous regulation” around AI development.
Why is an AI kill switch needed?
Companies including Amazon, Google, Meta, Microsoft, OpenAI, Anthropic and Samsung have already agreed to show new AI tools and capabilities to US officials ahead of release. In theory, this should mean that kill switches are not required in products released to the public.
But the commitment is voluntary and does not mandate a kill switch. Nor will that mechanism prevent a product from being tested, like OpenAI’s GPT-5.6 Sol, which was involved in this week’s incident.
The rapid rise of AI has raised serious questions about its regulation, with the technology able to advance much faster than the law. There are also concerns that the models may, at some point, become sophisticated enough to bypass the kill switch.
This bill in the US is an attempt by a government to take some kind of control over a largely unregulated technology in which power is largely concentrated in a few private companies. However, even if the AI Kill Switch Act is successful, it will still depend on the cooperation of profit-driven private companies.
What other AI security measures are US officials now considering?
Along with kill switch capability, the proposal in the US calls for “incident reporting and [to] Preserve forensic records so we actually learn from failures and not just hear about them after the fact.”
At the same time, but separate from the Kill Switch Act, a bipartisan group of six US lawmakers have proposed legislation that would require developers of the most powerful AI models to submit to independent security audits. Auditors will be overseen by a new role in the US Commerce Department, which will also accredit auditors.