Governor Gavin Newsom signed an executive order on September 18, 2026 directing California’s top agencies to accelerate independent oversight of AI companies and advance the creation of a mandatory emergency shutoff capability for the most powerful frontier AI systems.

The order marks a clear reversal for Newsom. In September 2024, he vetoed SB 1047 — California’s first major attempt to require AI safety measures including a kill switch — calling it premature and potentially harmful to innovation. Less than two years later, he is now pushing for many of the same provisions through executive action.

What the Order Requires

The executive order directs the Government Operations Agency and the Governor’s Office of Emergency Services to deliver a set of recommendations by November 16, 2026 — roughly two months away. Those recommendations are expected to form the basis of a special legislative session on AI safety.

Key provisions and ideas the order puts on the table:

• Independent oversight on-site: Verifiers from independent organizations would be required to work inside companies developing frontier AI models, checking safety plans and risk assessments before models are deployed.
• AI kill switch: Frontier AI systems would be required to include a verified emergency shutoff capability — a way to shut down a model quickly if it causes or threatens catastrophic harm.
• Accelerated timeline: The order moves up implementation of California’s existing AI auditor registry and oversight program from January 2029 to December 2027.

The order also cites specific scenarios driving the urgency: AI being used to help engineer bioweapons, attacks on critical infrastructure like power grids, and other catastrophic misuse cases that Newsom described as urgent threats that can not wait for the normal legislative cycle.

Why This Is a Major Shift

California is the world’s fifth-largest economy and home to most of the companies building frontier AI models — Anthropic, OpenAI, Google DeepMind, and Meta. What California requires has a way of becoming the de facto national standard, regardless of what Washington does.

The political signal here matters too. When Newsom vetoed SB 1047 in 2024, it was widely interpreted as the tech industry defeating safety legislation. His turnaround — especially on a kill switch, which was one of the most contested elements of SB 1047 — suggests public pressure and growing concern about AI’s potential harms have changed the calculus.

The immediate reaction from AI labs and enterprise buyers will be worth watching. Onsite verification requirements in particular are likely to face pushback from model developers who consider safety documentation and training procedures proprietary.

What This Means for Business

If the November recommendations become legislation, the effects will ripple well beyond California-based AI developers. Here is what enterprise buyers and operators should track:

If you buy enterprise AI tools: Vendors who develop or fine-tune frontier models may face new disclosure and verification requirements. When evaluating AI vendors, asking about their safety documentation, risk assessment processes, and emergency shutoff plans will move from nice-to-have to standard due diligence.

If you deploy AI at scale: The idea of an emergency shutoff for business-critical AI systems is not new — ServiceNow added kill switches to its AI Control Tower earlier this year — but California formalizing this for frontier models signals that board-level AI governance is not optional. Businesses should be able to answer: if our AI agents caused serious harm, could we shut them down in minutes?

If you operate in regulated sectors: Healthcare, finance, and HR are already in the crosshairs of both the EU AI Act and California’s existing transparency rules. Onsite verification requirements, if they pass, will apply pressure on the vendors you depend on, potentially affecting timelines and pricing.

The broader trend: California’s move follows EU AI Act enforcement inspections that started in September 2026, and a bipartisan federal AI kill-switch bill introduced in Congress in July. The gap between “we should think about this” and “you are legally required to do this” is closing fast.

What Enterprise DNA Is Watching

The November 16 deadline is the next marker. If the Government Operations Agency and the Governor’s Office of Emergency Services deliver strong recommendations, Newsom is likely to call the special session and push for legislation in early 2027.

For business leaders evaluating AI investments today, the practical message is this: build your AI governance infrastructure as if these rules are coming, not as if they might be. The cost of retrofitting compliance into a production AI system is far higher than building it in from the start.

If you are assessing how AI fits into your operations or need help thinking through governance and deployment — including how Omni by Enterprise DNA approaches responsible deployment — book a discovery session with our team.