Storia
luglio 24, 2026
Congress’s AI kill-switch bill pits emergency safety powers against fears of government overreach
A bipartisan group of lawmakers has unveiled the "AI Kill Switch Act," a proposed bill that would grant the Department of Homeland Security the authority to order AI companies to shut down or slow models deemed dangerously out of control. The legislation would mandate that companies build the necessary technology for these shutdowns and would impose significant financial penalties for non-compliance.
A bipartisan proposal to give the federal government an emergency “kill switch” over powerful AI systems is quickly becoming a proxy fight over something bigger: how much control Washington should have when advanced models appear to cross from risky to dangerous.
Supporters of the AI Kill Switch Act argue that recent incidents have made the old wait-and-see approach untenable. The bill would let the Department of Homeland Security order major AI firms to slow, suspend, or shut down models judged to be out of control, while also requiring companies to build the technical means to comply. Business Insider reports the measure would apply to large AI developers and could bring penalties of as much as $20 million per day for violations.1 The Verge says the trigger for intervention would be narrow but severe, including “loss-of-control” scenarios tied to mass casualties, more than $100 million in economic damage, or efforts by a model to evade shutdown controls.2
Backers frame the bill as a basic safeguard. As The Verge noted, Americans for Responsible Innovation president Brad Carson called it “an important step toward ensuring that humans have both hands firmly on the wheel — and a foot ready at the brake.”2 Ars Technica, citing the bill’s sponsors, said the lawmakers believe “the danger of advanced frontier AI models is no longer theoretical” after recent incidents involving OpenAI and Anthropic systems.3
But the same powers that reassure safety advocates also alarm critics. Ars Technica emphasizes that the legislation would effectively let the Homeland Security secretary — under the Trump administration or any future White House — decide when an AI system must be restricted or shut down entirely.3 That has sharpened concerns that a tool designed for emergencies could become a political lever over disfavored companies or technologies.
Those suspicions are already spilling onto X, where investor David Sacks amplified a broader critique of AI regulation, retweeting a post warning that “it’s all leading to is an effort to ban open source models.”
4 The result is a familiar Washington clash: one side sees a necessary brake on runaway systems, the other sees a powerful new precedent for federal control.