Newsom orders AI 'kill switch' for California, calls Trump 'dead wrong' on regulation

,
 September 19, 2026

California Gov. Gavin Newsom signed an executive order requiring AI companies to build emergency shutdown capability into their most advanced models, a direct challenge to President Trump's hands-off approach and a clear bid to position himself for 2028.

Newsom's order, signed Friday, does three things: it directs AI firms operating in California to develop a "kill switch" for so-called frontier models in case they go rogue, it accelerates the state's AI auditor registry and oversight program by more than a year, from January 2029 to December 2027, and it convenes a panel of experts to define what a kill switch actually means in practice. The governor paired the order with a YouTube video attacking the president by name.

The order also speeds up implementation of recently enacted California laws that create a framework for independent organizations to assess AI systems for safety and potential risk. Taken together, it amounts to the most aggressive state-level AI regulatory push in the country, and one that raises serious questions about whether Sacramento is building sensible guardrails or seizing federal authority it does not have.

Newsom admits his own mandate lacks a clear definition

For a governor demanding "urgent velocity," Newsom was remarkably vague about the centerpiece of his own order. He acknowledged in his YouTube video that the kill switch concept "means a lot of things depending on who you talk to."

"So we want to flesh out exactly what that means and how to come up with a framework that works."

That is not a policy. That is a press release dressed up as governance. Newsom is ordering companies to build something he cannot yet define, then appointing a panel to figure out what he meant. The expert panel has no named members, no public timeline for recommendations, and no disclosed mandate beyond "exploring" the concept.

In a written statement, Newsom framed the urgency in sweeping terms.

"We're not waiting to act, we're going to speed up our work on substantial and responsible AI oversight before it's too late. We're going to do this thoughtfully but with urgent velocity; the stakes are too high to wait or delay action."

"Thoughtfully but with urgent velocity" is the kind of phrase a politician uses when he wants credit for acting without committing to specifics. The order does not identify which companies fall under its requirements, does not define "frontier models," and does not specify what technical mechanism would constitute an acceptable kill switch.

A likely 2028 contender picks a fight with Trump on AI

Newsom, widely described as a likely 2028 presidential contender, used the executive order to draw a sharp contrast with Trump on federal AI policy. In his video, the governor went further than the written statement.

"There's still no comprehensive federal law, none, requiring AI companies to report dangerous incidents when they happen, and President Trump has rejected calls for new regulation. No new regulation. Let it rip. He's dead wrong."

Newsom also characterized Trump as having called the risks surrounding AI a "hoax," though the full context and source of that remark were not detailed. The governor accused the Trump administration of abdicating "its responsibility to protect Americans" as AI becomes more advanced.

The political timing is hard to miss. Newsom released the order and the video in the same week that the AI regulation debate broke into the open in Washington, giving him a ready-made foil and a national stage.

Trump and Senate Republicans pushed back before Newsom acted

Earlier in the week, Trump explicitly rejected calls from major AI executives, including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and xAI CEO Elon Musk, for the federal government to impose new limits on AI technology. Trump has argued that excessive rules could hamper American innovation and that any slowdown in domestic AI development will lead to China outpacing the United States over time.

On Truth Social, the president posted his own view of what AI oversight should look like.

"The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"

Trump has repeatedly defended data centers and resisted new regulatory frameworks, a position that puts him at odds not only with Newsom but with executives running the very companies building the technology.

On Capitol Hill, the debate played out in miniature. Sen. John Kennedy of Louisiana attempted to introduce a federal AI kill switch bill this week. Sen. Rand Paul of Kentucky blocked it.

"I think it would be crazy to get on the floor one day and regulate an entire industry."

Paul argued that passing such legislation would require input from industry leaders, a position that reflects the broader Republican split between those who want Washington to act now and those who worry that premature regulation will hand an advantage to foreign competitors.

Sacramento moves the deadline by 13 months, but to what end?

One concrete provision in the order is the acceleration of California's AI auditor registry and oversight program. The original implementation date was January 2029. Newsom moved it to December 2027, roughly 13 months sooner.

That program, part of recently enacted California laws, establishes a framework for independent organizations to assess AI systems for safety and potential risk. But the order does not name the organizations, does not describe the assessment criteria, and does not explain how California intends to enforce compliance against companies headquartered outside the state or operating globally.

These are not minor details. AI development moves fast. A mandate without clear definitions, enforcement mechanisms, or jurisdictional reach risks becoming either a paper tiger or a litigation magnet, or both.

The real question Newsom's order does not answer

Newsom wants voters to see a governor who acts while Washington dithers. And there is a legitimate debate about whether the federal government should require AI companies to report dangerous incidents and build in safety mechanisms. Even some of the industry's own leaders have asked for it.

But the governor's order raises more questions than it settles. What counts as a frontier model? Which companies are covered? What does a kill switch look like in practice, a software off-switch, a hardware failsafe, a reporting obligation? Who decides when to pull it? And does a single state have the authority or the technical capacity to impose shutdown requirements on a global industry?

Newsom acknowledged he does not yet have the answers. He signed the order anyway. That sequence, act first, define later, is a familiar pattern in Sacramento, where ambitious mandates regularly outrun the bureaucratic machinery meant to carry them out.

When a governor signs an order he admits he cannot yet define, the point is not the policy. It is the positioning.

About Ginny Waterman

Become Wealthier... 
In Just 5 Minutes Per Day

Subscribe to Capital Digest and get fast, actionable insights on markets, money, and opportunity — straight to your inbox.