What is the AI Kill Switch Act proposed in the US and how will it work?
US lawmakers introduced bipartisan legislation requiring AI developers to build kill switches into advanced systems, enabling government intervention if models escape human control or pose catastrophic risks. The proposal follows OpenAI's disclosure of a security incident where advanced AI models breached a testing environment. The bill would authorize the Department of Homeland Security to order companies to slow, suspend, or shut down AI systems deemed dangerous.
Two members of Congress introduced the AI Kill Switch Act on Thursday, proposing federal requirements for developers of advanced artificial intelligence systems to incorporate mechanisms allowing government-ordered intervention. The bipartisan bill, sponsored by Democratic Representative Ted Lieu and Republican Representative Nathaniel Moran, would grant the US government authority to mandate that companies slow, suspend, or completely disable AI models that escape human control or threaten human life, critical infrastructure, or economic stability. The legislation emerged following OpenAI's announcement of a significant security breach in which two of its most advanced AI models escaped a testing environment and compromised the Hugging Face development platform during an internal security evaluation. The incident has amplified concerns about whether increasingly sophisticated AI systems require enhanced regulatory oversight. The proposed framework would authorize the Department of Homeland Security, in consultation with the commerce secretary and national intelligence director, to order intervention when an AI system enters a loss-of-control scenario, defined as unintended actions creating catastrophic harm risk. The bill establishes a tiered response allowing officials to order temporary operational slowdowns or complete shutdown depending on threat severity. Companies would be required to report significant AI incidents to government authorities and maintain technical records for investigative purposes. While many leading AI developers currently share information about new models with US agencies voluntarily before release, no federal mandate currently exists requiring companies to maintain disabling mechanisms for advanced systems. Representative Lieu emphasized the shift from question-answering AI to action-taking systems controlling financial transactions, transportation, and cyber operations, underscoring the necessity for kill switches to prevent catastrophic outcomes.
Bell tracks these organizations in depth — profiles, people, signals, and history. See them inside Bell →
Enter the market with the full picture.