Situation report: California is moving on AI oversight while Washington sits on its hands. Governor Gavin Newsom signed an executive order Friday that could lead to a mandated “kill switch” for frontier AI models, according to The Verge AI. The order gives a panel of experts two months to deliver recommendations on hardening AI safety rules in state law.
Threat framing is straightforward. The governor is betting that federal action isn’t coming, so the state builds the framework itself and dares Congress to adopt it.
What the order actually does
The Verge AI reports the expert group has a specific checklist to work through. Newsom wants them to weigh four mandates for AI companies:
- Embed independent verification teams onsite for regular audits.
- Subject transparency reports and risk assessments to independent auditor standards.
- Build a “kill switch” that gets routinely tested to confirm it works.
- Report “loss-of-control incidents” as critical safety events. The article cites the OpenAI attack on Hugging Face as the type of incident that would qualify.
The order also tells a state agency to speed up rollout of two laws Newsom already signed: one creates a framework for independent verifiers to assess AI safety, the other sets up a state registry of AI auditors.
So this isn’t a blank-slate proposal. It’s an acceleration of machinery that already exists on paper.
Why the timing matters
The federal picture is bleak for anyone hoping for national rules. As detailed in The Verge AI, Congress has been buzzing with bill discussions since former Anthropic researcher Jacob Coxon posted a warning that there’s a more than 10 percent chance AI “could kill all humans” within a decade. Buzz is cheap. Votes aren’t.
The House is on recess until November. Midterms make any major Senate move unlikely. And the White House has staked out a clear position. President Trump has insisted, “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!”
Newsom’s response cuts hard. “The federal government’s abject failure to create any form of meaningful AI oversight or accountability should alarm every American, especially when AI CEOs themselves are begging for regulation,” he said in a statement. He added, “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.”
The California-as-floor strategy
What stands out here is the framing. Newsom isn’t just regulating his state. He’s explicitly calling on Congress and Trump “to review and adopt the state’s framework, or use its nation-leading regulation as a floor, not a ceiling, for the benefit and safety of all Americans.”
This is the California playbook. Emissions standards. Privacy law. Set the bar in the state where the companies live, and the rest of the country inherits it by default. Nearly every frontier lab is headquartered in California, so state law effectively becomes industry law.
And more is coming. The Verge AI notes Newsom told Politico on Wednesday he may call lawmakers into a special legislative session on AI, alongside further executive action.
Tactical implications for practitioners
If you build on or deploy frontier models, here’s what to watch:
- Audit exposure. Onsite independent verification would be a step change from today’s self-reported safety cards. Expect labs to start hiring compliance staff and pushing back on scope.
- Kill switch engineering. A verified, routinely tested shutdown mechanism is a real technical requirement, not a slogan. How it works for models already distributed via API or open weights is an open question the panel will have to answer.
- Incident reporting. A formal “loss-of-control” category means labs may have to disclose events they’d currently keep quiet. That’s good for transparency and uncomfortable for PR teams.
- Timeline. Two months for recommendations. Then possibly a special session. This could become law faster than anything in Washington.
My read: the kill switch grabs headlines, but the audit and incident-reporting provisions will reshape how labs operate day to day. A shutdown button is only useful if someone independent has verified it exists.
The panel’s recommendations land in roughly 60 days. Until then, the federal government has a choice. Adopt California’s floor, or watch the country’s AI rules get written in Sacramento anyway. The full report from The Verge AI has the complete statement and order details.