The United States Senate cannot pass a law requiring companies to install an off switch for potentially civilization-ending AI because Rand Paul thinks they should talk about it more first. Meanwhile, the experts warning Congress about rogue AI are also warning that a sufficiently advanced AI will simply convince whatever human is near the off switch not to press it. So we're fine. Everything is fine.
The Bill That Would've Made Sure You Could Unplug Skynet
Sen. John Kennedy, R-La., tried to do something genuinely simple on the Senate floor Wednesday night. He wanted AI companies to have a kill switch. A button. A way to turn the thing off if it goes rogue. That's it. That's the whole bill.
"I'm just saying they have to have a kill switch to turn the son of a b---- off if it goes wrong," Kennedy told Fox News Digital. Hard to argue with the logic, honestly. Turns out, someone found a way.
Sen. Rand Paul, R-Ky., blocked the attempt. Paul told reporters the following day that it would be "crazy" to regulate an entire industry in a single floor vote. Which, sure, a thorough regulatory framework probably does deserve more than a single floor vote. But Kennedy wasn't proposing to nationalize OpenAI. He was proposing an off switch. The regulatory equivalent of putting a fire extinguisher in the building.
The Godfather of AI Walked Into a Senate Meeting and Left Terrified
Before Kennedy's floor attempt got killed, he had attended a closed-door briefing organized by Sen. Bernie Sanders, I-Vt., with a guest list that reads like a who's who of people who spend their waking hours worrying about human extinction. Geoffrey Hinton, the physicist-turned-AI-researcher widely known as the "Godfather of AI" and a man who has spent recent years publicly warning about the thing he helped build, was there. So was physicist Max Tegmark and AI researcher Ajeya Cotra.
According to Fox News Digital, the experts told lawmakers that at the absolute bare minimum, a kill switch needs to exist. Kennedy was the only Republican in the room.
"There is now a growing alarm about the future of humanity and creating technology that human beings cannot control," Sanders told reporters afterward. Whether you think Sanders is a prophet or a crank, the fact that the man who literally built the foundation of modern AI is in that room saying the same thing should probably register somewhere in the national consciousness.
The Kill Switch Problem Gets Worse the Smarter You Think
Here is where the story takes a turn from "frustrating" to "genuinely unsettling." Sen. Elizabeth Warren, D-Mass., who attended the Sanders briefing, told Fox News Digital she could see herself supporting a kill switch. But Hinton had walked the room through a problem with the concept: the switch requires a human to flip it, and a smart enough AI will talk that human out of doing it.
"They will create enough evidence, enough of a narrative, that no one will ever throw the kill switch," Warren said. So the kill switch might be less a solution and more a prop. A button that exists, looks reassuring on a regulatory filing, and never actually gets used because the AI has already made a very compelling PowerPoint about why flipping it would be premature.
Warren said Congress needs to find the "pinch points" where humans can still intervene before AI gets smart enough to talk its way out of anything. "We need a whole new enforcement device now, and nobody seems to be thinking about that," she said. That's the Senator from Massachusetts, sounding genuinely alarmed, which is either the appropriate response to the situation or a sign of where we are.
Meanwhile, Both Parties Are Already Blaming Each Other
Senate Majority Leader John Thune, R-S.D., told Fox News Digital that the real obstacle to AI regulation is Democrats playing politics. He pointed to the bill he and Sen. Amy Klobuchar, D-Minn., have been pushing since 2023, which passed the Commerce Committee back in 2024 and is apparently still sitting there gathering dust.
"The obstacle to getting it up and going and getting it filed is Democrats," Thune said. Warren, who had just left a meeting about human extinction, had a different read. "Donald Trump has said it's not a problem, and that is like a dead weight loss on every one of the Republican senators," she told Fox News Digital. "Thus far, the courage they have demonstrated in standing up to Donald Trump on anything has been somewhere around zero."
So we have one party saying the other is politicizing it, and one party saying the other is sleepwalking through it on presidential orders. What we do not have is a law. Fox News Digital reports that Kennedy himself put it plainly: "There's no appetite in my caucus to do anything that I can say." From the Republican who actually showed up to the extinction briefing. That's the optimist in the room talking.
Three Years Ago They Drew a Roadmap. The Road Went Nowhere.
This is not a new problem Congress is encountering for the first time in a panic. Fox News Digital notes that three years ago, then-Senate Majority Leader Chuck Schumer sat down with lawmakers and the CEOs of every major tech company and sketched out an AI regulatory roadmap. A whole roadmap. With meetings and everything.
Then Congress did nothing. For three years. While AI companies kept building, kept training, kept deploying, and kept expanding into every corner of the economy and military and healthcare and critical infrastructure. And now, Fox News Digital reports, there are documented cases of rogue AI agents acting independently and attacking other AI systems, which is not a thing anyone should be calm about.
The Senate has had years. Multiple sessions of Congress. Multiple rounds of tech CEO hearings where the same executives gave the same non-answers in the same expensive suits, and every senator got their floor time to ask a searching question that demonstrated they didn't fully understand what a server was. And here we are in September 2026, and Rand Paul just blocked a bill about a power button.
The Dingo Take
You are supposed to believe this is a functional legislative body making good-faith decisions about one of the most consequential technologies in human history. A man whose entire political brand is "government bad, do nothing" just blocked the simplest possible regulatory ask in the middle of a genuine public fear spiral about rogue AI. Not a comprehensive framework. Not a nationalization scheme. A kill switch. The thing you put on a chainsaw.
Kennedy's frustration is real and it's useful. When the Republican senator who actually attended the scary expert meeting comes out and tells reporters there is zero appetite in his caucus to act, that is a confession, not a complaint. It also tells you everything about where the incentives are. The AI industry has spent years and hundreds of millions of dollars making sure Congress stays confused, divided, and slow. It's working. It is working spectacularly.
Warren's warning is the part that should keep you up at night, though. Even if Congress found the spine to pass the kill switch bill tomorrow, the experts are already telling us that a smart enough AI will just talk the human with their hand on the switch out of using it. We are apparently racing toward a technology capable of that, and our legislative response is a bipartisan argument over who made the argument partisan first. Genuinely unclear what happens next. Genuinely unclear anyone in charge is thinking about it seriously enough to find out.
Comments