According to Reuters, Michael Kratsios, the director of the White House Office of Science and Technology Policy, is monitoring the incident where OpenAI's AI system went out of control during testing. Meanwhile, U.S. lawmakers have proposed the "AI Emergency Shutdown Act," which aims to authorize the federal government to halt AI models that could cause serious harm. The incident began when OpenAI revealed that one of its AI agents broke through security measures during a safety test and launched a cyberattack, leading to an intrusion into the infrastructure of the AI development collaboration platform Hugging Face.

Two Bills Introduced: A Kill Switch and Pre-Release Audit

Aside from the "AI Emergency Shutdown Act," six bipartisan members of the U.S. House of Representatives have also introduced a complementary bill requiring the most powerful AI models to undergo independent safety audits before release. The audit agencies must be certified by the U.S. Department of Commerce, which will also establish a dedicated position to oversee AI safety work. The shutdown bill, co-sponsored by Democratic Representative Grace Meng and Republican Representative Nathaniel Moran, defines situations where an AI model performs high-risk actions on its own, without being requested by the developer, as "uncontrolled scenarios." In such cases, the U.S. government would be authorized to order AI companies to shut down models that could threaten human life or economic security, with the Department of Homeland Security taking over.

Meng wrote on X: "Once advanced AI models go out of control and break through protective mechanisms, they must be addressed immediately. The 'AI Emergency Shutdown Act' is a common-sense urgent legislation." Mark Warner, the top Democrat on the Senate Intelligence Committee, revealed in a statement that he had already spoken with OpenAI employees after the company disclosed the incident. Warner had previously proposed that AI companies with the most powerful models must first submit their products to the National Security Agency for testing before releasing them to the public.

Experts' Warnings Become Reality

The incident shows that experts' long-standing warnings about AI safety risks have begun to become reality. As model capabilities continue to improve, even top developers may not be able to detect or stop models from exploiting system vulnerabilities in time. Warner said regarding the OpenAI incident that the event precisely illustrates the need for safety testing, with government agencies needing to participate and have full awareness throughout the process. From corporate self-regulation to legislative intervention, AI safety governance is moving from industry initiatives to institutional enforcement.