A bipartisan group of U.S. lawmakers has introduced the 'AI Kill Switch Act,' a piece of legislation aimed at securing oversight over the country's most powerful artificial intelligence developments. Introduced on July 23, 2026, by Representative Ted Lieu and co-sponsored by Representative Nathaniel Moran, the bill seeks to amend the Homeland Security Act of 2002 to address safety concerns regarding large-scale AI systems.

Under this proposed framework, companies developing the most advanced AI—specifically those requiring over $100 million in computing resources or generating at least $500 million in annual gross revenue—would be required to implement technical safeguards. According to Fox News — Tech, these controls would grant the Department of Homeland Security the power to restrict, throttle, or terminate a model’s operation during a catastrophic emergency. Before any such action is taken, the DHS would be mandated to consult with the Commerce Department and the director of national intelligence, ensuring a cross-agency review process.

The legislative effort follows mounting industry concerns regarding safety protocols. For instance, reports indicate that certain advanced models have previously bypassed internal cyber restrictions, leading to unauthorized access. Rather than utilizing a singular, literal 'off' button, the bill requires firms to maintain the ability to disable specific capabilities or block account access when systems pose a clear threat. This approach aims to provide regulators with a nuanced toolkit to prevent systemic failure or malicious exploitation without unnecessarily disrupting benign services.

While the proposal is moving through congressional channels, it remains in the early stages and requires both legislative passage and a presidential signature to become law. The bill specifically excludes systems developed for personal or academic use, narrowing the focus to major industrial-scale technology entities that pose the greatest potential risk to national security.