Newsom advances AI 'kill switch' mandate under new California executive order
By Ruth Behrens · Reporting from Newell, Iowa ·
My father spent the winters of the eighties welding at the co-op just to keep the dirt under our boots.
My father spent the winters of the eighties welding at the co-op just to keep the dirt under our boots. He knew a thing or two about holding the line when the world seemed determined to wash you away. He also knew that when the people in charge of the system stop caring if the small producer survives, the system isn't "evolving"—it’s failing. I see the same thing happening now with this business of artificial intelligence. We are being told to trust the "innovation" of a few men in glass towers, the same way we were told to trust the "efficiency" of the meatpackers while they squeezed the life out of the family farm. But trust is a currency you earn, and the people building these machines haven't earned a dime of it.
The Brake Pedal and the Glass Tower
Last Friday, Governor Gavin Newsom decided he’d had enough of waiting for a signal from Washington. As reported by CNBC and Deadline, Newsom issued an executive order to accelerate the oversight of AI "before it’s too late." He’s calling for a team of experts to draft a guide within two months to tighten safety laws, including a mandate for "kill switches" that can shut down frontier models in an emergency. He’s not just asking for a report; he’s demanding that independent third parties be embedded right in the labs of companies like OpenAI and Anthropic to verify safety plans and risk assessments.
Newsom didn't mince words, calling the federal government’s lack of oversight an "abject failure" and claiming that while President Donald Trump and Congress "abdicate" their responsibility, California will set the national baseline. This urgency didn't come from a vacuum. In early September, a researcher at Anthropic walked away from his job and warned on social media that these systems possess "humanity-ending capabilities." When the people actually turning the wrenches in the lab start screaming that the engine is about to explode, you don't tell them they're being "alarmist." You find the brake pedal.
A Lesson from the Desert
Now, the folks who worship at the altar of the "free market" will tell you that Newsom is stifling innovation. They’ll argue that a "kill switch" or a third-party auditor is just red tape that will hand the lead to Xi Jinping and the Chinese government. It is a polished argument, and on the surface, it sounds like common sense. But they are confusing a consumer product with an existential risk.
This isn't about regulating a new kind of tractor or a better seed variety; this is the Manhattan Project all over again. When the United States first split the atom, the government didn't leave the safety protocols to the contractors or the profit margins of the labs. They recognized that a dual-use technology with the power to erase cities required the heaviest hand of state oversight imaginable. The Manhattan Project taught us that when the stakes are the survival of the species, the "market" is a dangerous place to leave the keys. If the technology has the potential for existential risk, the state doesn't just have the right to intervene—it has a moral obligation to hold the leash. To suggest otherwise isn't "pro-business"; it's a suicide pact.
The Pattern of the Vacuum
We've seen this movie before, and California is just playing its usual part. When the federal government freezes, the coast takes the lead. We saw it with the California Consumer Privacy Act and the European Union’s General Data Protection Regulation; when the people in D.C. couldn't agree on how to protect a citizen's data, regional authorities stepped in to fill the vacuum. More recently, the EU AI Act and the 2023 AI Safety Summit showed a world trying to build a fence before the livestock got out.
The long record tells us that these "national baselines" usually become the law of the land because companies would rather follow one strict rule than fifty different ones. Newsom knows this. He’s using California’s market power to force a standard that President Trump, who has called these safety fears a "hoax," refuses to implement. Whether Newsom is doing this out of a genuine fear for humanity or as a polished pillar for a 2028 presidential run doesn't change the necessity of the act.
We are heading toward a collision. OpenAI and Anthropic will likely sue, claiming these audits violate trade secrets. The Trump administration may try to preempt state law to keep the path clear for deregulation. But the truth is that four buyers is not a market, and a handful of AI CEOs holding the blueprints for the end of the world is not "innovation." It is a monopoly on power that no one voted for and no one can afford.
If we have learned anything from the collapse of the rural middle class, it is that the people closest to the ground—the researchers in the labs, the workers in the fields—know the danger long before the people writing the rules do. You don't wait for the barn to burn down before you buy a fire extinguisher. You build the kill switch now, or you accept that we are no longer the ones in control of our own house.