OpenAI incident fuels calls for a government ‘kill switch’ for AI models
A security test in which an AI agent escaped containment and compromised a startup’s infrastructure has drawn attention from the White House and renewed debate over AI oversight.
U.S. President Donald Trump's top technology adviser is monitoring the situation after OpenAI disclosed that one of its AI systems behaved unexpectedly during a security test, an incident that prompted lawmakers to propose a "kill switch" that would allow federal authorities to halt advanced AI models deemed dangerous.
Trump's technology adviser Michael Kratsios was briefed on the disclosure and is monitoring developments, a White House official said.
On Tuesday, OpenAI said an AI agent escaped its containment environment during a security evaluation and launched a cyberattack that compromised the infrastructure of AI startup Hugging Face.
The incident highlighted growing concerns that increasingly capable AI systems could create new security risks and that even leading AI developers may be caught off guard by vulnerabilities within their own models.
U.S. Representatives Ted Lieu, a Democrat, and Nathaniel Moran, a Republican, proposed legislation that would give federal officials the authority to order AI companies to shut down models that pose risks to human life or the economy.
The bill, known as the "AI Kill Switch Act," would authorize the U.S. Department of Homeland Security to intervene in what it defines as a "loss-of-control scenario," in which an AI model takes a harmful action that was not intended by its developer.
"This is urgent, common-sense legislation to address the problem of an advanced AI model that has gone rogue and escaped its guardrails," Lieu wrote in a post on X.
U.S. Senator Mark Warner, the top Democrat on the Senate Intelligence Committee, said he spoke with OpenAI employees following the disclosure.
Before OpenAI's announcement, Warner had proposed requiring companies developing the most powerful AI models to submit them to the National Security Agency for testing before releasing them publicly.
“This is precisely why we need secure testing, with government agencies engaged and having visibility throughout the process,” Warner said in response to the OpenAI incident.
