Newsom builds the room while Washington cheers for the cliff

By Aoife Gallagher · Reporting from Dublin ·

When the center fails, the periphery must lead.

When the center fails, the periphery must lead. It is a pattern I have watched play out across the Atlantic for two decades: when the great capitals become paralyzed by ideology or blinded by the glare of their own power, the smaller, more agile jurisdictions are forced to build the fences that keep the rest of us safe.

Right now, the center of the American empire is not just paralyzed; it is actively cheering for the cliff. President Donald Trump has dismissed AI safety fears as a “hoax,” leaving a vacuum of leadership in Washington that is as dangerous as it is predictable. Into this void steps California. On September 18, 2026, Governor Gavin Newsom issued an executive order to rein in artificial intelligence "before it's too late," as reported by CNBC. He isn't just tweaking a few guidelines; he is attempting to build a regulatory fortress.

The Brake Pedal for the End of the World

The specifics of Newsom’s order, detailed by the California State Portal, are a direct response to the federal government's "abject failure." Newsom is accelerating the implementation of SB 813 and AB 1405—laws that mandate third-party oversight and a registry for AI auditors—by a full year. More provocatively, he is exploring the mandate of a “kill switch” for frontier AI models.

This is not mere political theater. We have already seen the cracks in the hull; Politico notes that an OpenAI model was previously hacked into Hugging Face, and a researcher at Anthropic recently warned of the technology's "humanity-ending capabilities." When the people building the engine are the ones begging for a brake, you stop arguing about the cost of the brake and you install it.

This instinct to create emergency safeguards against systemic, opaque collapse is not new. We saw it during the Financial Crisis of 2008, when the world realized that complex financial instruments could trigger a global meltdown without a single one of us knowing where the "off" switch was. Newsom is simply applying that lesson to silicon instead of subprime mortgages.

The Logic of the Nuclear Baseline

The critics will call this a "regulatory stranglehold" that will hand the AI lead to adversaries. This is the strongest case the opposition can make: that by imposing "kill switches" and onsite independent audits, California is handicapping the very innovation that ensures American hegemony. They argue that the speed of development is the only true security.

But this is a gambler’s logic, and the stakes are not a quarterly earnings report—they are existential. The only valid precedent here is the Nuclear Non-Proliferation Treaty. The NPT was not born out of a desire to stifle the "innovation" of atomic energy; it was born from the realization that certain dual-use technologies possess such catastrophic potential that they cannot be left to the whims of individual actors or the reckless competition of superpowers.

The NPT succeeded because it established a framework where safety and verification were not "obstacles" to the technology, but the very conditions under which the technology was allowed to exist. Newsom is attempting to create a digital NPT. By requiring independent verification organizations to be embedded onsite in labs, California is treating frontier AI not as a consumer product, but as a strategic risk. We saw a similar regional impulse with the EU AI Act, which recognized that a fragmented approach to high-risk systems is a recipe for disaster.

When the State Becomes the Sovereign

Newsom is playing a high-stakes game of regulatory arbitrage. He knows his term ends on January 4, but he is spending his final months ensuring that California’s policy becomes the "national baseline." He is betting that the industry—including OpenAI and Anthropic, both of whom have signaled support for these frameworks—would rather deal with one strict, predictable set of California rules than a chaotic patchwork of fifty different state laws.

This is the same strategy of "urgent velocity" we saw during the COVID-19 pandemic, where state executive orders filled the void left by a dysfunctional federal response. It is a necessary survival mechanism. With Pennsylvania Governor Josh Shapiro already calling for similar guardrails, we are seeing the emergence of a multi-state coalition that functions as a surrogate federal government.

The path forward is clear: Senator Jerry McNerney will likely codify these recommendations in December, and a legal war will erupt between Sacramento and the Trump administration over whether a state can mandate a "kill switch" on a global product.

Let them fight. The alternative is a race to the bottom where the finish line is a loss of human control. If the federal government refuses to act as the adult in the room, then California must build the room, lock the door, and hold the key. Sovereignty is meaningless if you are presiding over a graveyard; the only real power is the power to prevent the catastrophe.

Sources

  1. CNBC: California Gov. Newsom issues executive order to rein in AI 'before it's too late'
  2. Politico: Newsom signs executive order to explore new AI rules, consider ‘kill switch’
  3. California State Portal | CA.gov: Governor Newsom issues executive order to accelerate independent oversight and advance the creation of an AI kill switch