The Sovereign Off-Switch: California's Gambit for AI Control

When we look at the horizon of artificial intelligence, the central tension isn't just about whether the tech works, but who holds the power to stop it. We are witnessing the first attempts to codify a sovereign off-switch for a form of intelligence that may eventually outpace the state's ability to govern it through traditional legislation.

According to reporting from The Verge, California Governor Gavin Newsom (D) is positioning his state to lead this charge. In an executive order issued Friday, Newsom directed the state to assemble a group of experts tasked with delivering recommendations within two months on strengthening AI safety measures in state law.

***

**Opinion: The Power Play**

While the official rhetoric focuses on "safety," the insistence on a "kill switch" that is routinely verified as effective suggests something deeper. This isn't just a safety feature; it is an attempt to establish a hard-coded point of leverage. By mandating that frontier models have a verifiable off-switch, California is attempting to ensure that the state retains ultimate authority over systems that could otherwise operate beyond the reach of bureaucratic oversight.

***

As detailed by The Verge, Newsom's proposed framework includes several aggressive oversight mechanisms. He wants the expert group to consider requiring AI companies to:

* Place independent verification teams onsite to conduct regular audits. * Submit transparency reports and risk assessments to the standards of independent auditors. * Implement a "kill switch" that is regularly verified for effectiveness. * Report "loss-of-control incidents"—citing the OpenAI attack on Hugging Face as a critical safety incident—as mandatory disclosures.

Newsom is also directing a state agency to accelerate the implementation of two previously signed laws that establish a state registry of AI auditors and a framework for independent verifiers to assess AI safety.

This urgency comes at a time of federal paralysis. The Verge reports that while former Anthropic researcher Jacob Coxon has warned of a greater than 10 percent chance that AI could kill all humans within a decade, Congress appears unlikely to act before the midterms. The House is currently on recess until November, and Senate movement is stalled by the upcoming elections.

Contrastingly, President Donald Trump has dismissed the need for such guardrails, stating that the only control AI needs is a "STRONG AND SMART (High IQ!) PRESIDENT," which he claims the U.S. already possesses.

Newsom has been explicit about his intentions. In a statement, he called the federal government's lack of meaningful oversight an "abject failure" and noted that AI CEOs themselves are requesting regulation. In an interview with Politico, Newsom floated the possibility of calling California lawmakers back for a special legislative session. He is positioning California's framework as a model for the rest of the country, urging Congress and Trump to adopt it as a "floor, not a ceiling."