Republican Senator Pushes Bill to Force AI Companies to Build Emergency Shutdown Switches

A New Kind of Safety Measure for a New Kind of Threat

Sen. John Kennedy of Louisiana introduced legislation Wednesday requiring artificial intelligence models to have a "kill switch" — a mechanism that would allow companies to completely shut down any AI model if it acts amorally or immorally outside of human control. The move marks one of the most direct attempts yet by a sitting U.S. senator to force the AI industry to build hard safety limits into its own products.

Kennedy framed the threat in stark terms, describing AI as "a brand new species that acts like a human, but without most of the values that decent humans share," warning that through recursive self-improvement, AI could start acting on its own in an amoral or immoral way. His answer, he said, is to require all AI models to have a kill switch.

How the Bill Would Work — and What It Won't Do

Kennedy introduced the bill under unanimous consent on Wednesday, explaining it was "the only way you're going to get anything done" given the constraints facing the Senate over the next several weeks. Unanimous consent is a procedural shortcut that allows legislation to pass without a full floor vote — but it also means a single senator can block it.

The bill mirrors similar legislation introduced in the House in July by Reps. Ted Lieu (D-Calif.) and Nathaniel Moran (R-Texas), which would allow the government to order top AI firms to shut down or slow AI models deemed too dangerous and require companies to report incidents. But Kennedy's version takes a notably different approach to who holds the power. Kennedy stressed his bill would not permit the government to halt these powerful AI systems, explaining that putting the federal government in charge of the kill switch would doom the bill's chances — and likely draw a presidential veto, given that the president is "not too keen on AI regulation."

The proposed measure is expected to specifically target developers of frontier AI models and would require companies to build failsafes directly into their core architecture. As of now, there are no cosponsors for the bill.

What Triggered the Urgency

As recently as the start of September, Congress was hesitant to address public concerns around AI's rapid development. That changed when researchers warned in social media posts that AI posed an existential threat to humanity, and top industry executives backed calls for restraint.

Evan Hubinger, who leads Anthropic's efforts to align its models' values with humans', posted on X that there is a 10 percent chance AI could kill all people within the next decade. Anthropic CEO Dario Amodei also published a letter arguing AI could outrun humans' ability to understand or control what they were creating. The alarms from inside the industry itself appear to have been the tipping point for lawmakers who had previously been reluctant to act.

A Congress Divided on How Far to Go

President Trump responded by calling the warnings against AI a "sick conspiracy" and argued no guardrails were needed. Not everyone in his party agreed. Many lawmakers, including Senate Majority Leader John Thune, Speaker Mike Johnson, and Rep. Tim Burchett, said guardrails needed to be put in place.

House Majority Leader Steve Scalise and Republican Ohio Sen. Bernie Moreno told the Daily Caller News Foundation that Trump rightly highlighted the competitive race with China over AI and what they called the "irrational fears" about the technology. That tension — between those demanding safety mechanisms and those worried about stifling American innovation — will likely define the debate as Kennedy's bill moves forward. Whether a kill switch becomes law or simply a rallying point, it signals that Washington's long era of AI hands-off governance may finally be coming to an end.