Governor Gavin Newsom on September 18 signed an executive order directing the acceleration of independent oversight mechanisms for artificial intelligence systems, including the development of an AI “kill switch” capability designed to allow human operators to shut down AI systems that exhibit dangerous or uncontrolled behavior. The order also calls for national adoption of California’s AI regulatory framework, a direct challenge to the Trump administration’s position that the federal government should preempt state-level AI regulation. California is home to 33 of the top 50 privately held AI companies in the world and commands the fourth-largest economy on the planet, giving the state outsized leverage over an industry that is headquartered largely within its borders.
Key Takeaways
- Governor Newsom signed an executive order on September 18 directing the acceleration of AI oversight mechanisms, including a “kill switch” capability for AI systems
- The order calls for national adoption of California’s AI regulatory framework, positioning the state as the model for how the country should govern artificial intelligence
- California is home to 33 of the top 50 privately held AI companies globally and a quarter of all AI patents, conference papers, and companies worldwide
- The executive order builds on Newsom’s March 2026 order (N-5-26), which required AI vendors seeking state contracts to demonstrate safeguards against bias, illegal content, and civil rights violations
- The Trump administration’s AI policy framework, released earlier in 2026, urged Congress to preempt state AI regulations, calling them “burdensome” to national competitiveness
- Union leaders have publicly stated they will not support Newsom’s presidential ambitions without stronger worker protections from AI displacement
What the Executive Order Actually Directs
The September 18 executive order operates on two tracks. The first is operational: it directs California state agencies to accelerate the development of independent oversight mechanisms for AI systems used in or contracted by the state. The “kill switch” concept, which has circulated in AI safety research circles for years, refers to the ability of a human operator or an automated monitoring system to shut down an AI model or agent that is operating outside its intended parameters. The order does not define the precise technical specifications of such a mechanism but directs agencies to move faster on building or procuring one.
The second track is political. The order explicitly calls for the national adoption of California’s AI framework, a statement that carries weight because of the state’s position as the center of the global AI industry. When California sets procurement standards for AI vendors, those standards do not stay within state lines. Companies that want to do business with California’s government adapt their products and policies to meet California’s requirements, and those adaptations frequently become the default for how the same companies operate in other markets. The dynamic is familiar from California’s role in shaping automobile emissions standards, consumer privacy law through the California Consumer Privacy Act (CCPA), and energy policy. The AI executive order follows the same playbook: set the standard in California, and the industry will carry it everywhere else because building separate products for separate regulatory regimes is more expensive than building to the highest bar.
The March Executive Order That Set the Foundation
The September order does not stand alone. It builds on Executive Order N-5-26, which Newsom signed on March 30, 2026, and which established the procurement-side regulatory framework that the September order now seeks to accelerate and expand. The March order directed the California Department of General Services and the Department of Technology to develop new certification requirements for AI companies seeking contracts with the state. Within 120 days, those agencies were required to submit recommendations for standards that would require vendors to attest to and explain their policies on several specific issues.
Those issues include the distribution of illegal content generated or facilitated by AI systems, harmful model bias that produces discriminatory outputs, and violations of civil rights and civil liberties including free speech, voting rights, and protections from unlawful surveillance. The March order also required the Department of Technology to issue guidance on watermarking AI-generated or significantly manipulated images and video, addressing the deepfake and synthetic media problem that has become a persistent concern in elections, journalism, and public discourse.
The 120-day deadline from the March order fell on July 28, 2026. The recommendations submitted by the two agencies form the basis for the procurement standards that are now being implemented across state contracting processes. The September order pushes those standards further by adding the kill switch requirement and by framing California’s approach as a national model rather than a state-level experiment.
The Federal Government Wants to Go in the Opposite Direction
The tension between Sacramento and Washington on AI regulation has been building throughout 2026. The Trump administration released a national AI policy framework earlier this year that laid out legislative recommendations to Congress. The framework’s central argument is that AI development is “an inherently interstate phenomenon with key foreign policy and national security implications” and that states should not be permitted to regulate it independently. The document called for preempting state AI laws that the administration considers “burdensome” and advocated for a light-touch regulatory approach that prioritizes innovation speed over safety requirements.
Newsom’s office has responded to the federal framework in pointed terms. “Unlike the Trump administration, California remains committed to ensuring that AI solutions adopted and deployed by the state cannot be misused by bad actors,” the Governor’s Office said in a press release accompanying the March order. The September order doubles down on that position by explicitly calling for other states and the federal government to adopt California’s approach rather than the White House’s deregulatory model.
The disagreement is not abstract. It determines what rules AI companies follow when they build products, what disclosures they make about how those products work, what safeguards they implement against bias and misuse, and whether any government entity has the authority to require a company to shut down an AI system that is causing harm. California’s position is that the answer to all of those questions should involve binding requirements enforced at the state level. The Trump administration’s position is that those requirements should be minimal and set exclusively by the federal government.
The Industry Newsom Is Trying to Regulate Operates in His Own State
The political complexity of Newsom’s position is that the industry he is regulating is also the economic engine of the state he governs. California is home to 33 of the top 50 privately held AI companies in the world, according to the Governor’s Office. The state holds a quarter of all AI patents globally. San Francisco alone has the highest concentration of generative AI job postings and the highest number of well-funded AI startups of any city in the country. The AI industry’s presence in California generates tax revenue, creates high-paying jobs, and attracts venture capital that flows through the state’s economy.
Regulating that industry aggressively carries a risk that companies relocate to states or countries with lighter regulatory environments. That risk is not theoretical. AI companies have already established operations in Texas, Florida, and internationally in jurisdictions that have signaled a more permissive approach. But the concentration of talent, infrastructure, and institutional knowledge in the San Francisco Bay Area gives California a gravitational pull that is difficult to replicate elsewhere. The state’s regulatory leverage is strongest when companies calculate that the cost of compliance is lower than the cost of leaving.
Newsom has navigated this tension before. In 2024, he vetoed SB 1047, a landmark AI safety bill authored by State Senator Scott Wiener that would have imposed broad liability requirements on AI developers. Newsom argued at the time that the bill was too blunt and could stifle innovation. The veto drew criticism from safety advocates who saw it as evidence that the Governor was unwilling to confront the industry directly. The executive order approach allows Newsom to impose requirements through procurement standards rather than legislation, which gives him more control over scope and timing and avoids the legislative negotiation process that produced the provisions he objected to in SB 1047.
Labor Pressure Is Shaping the Timeline
The September executive order also arrives under pressure from organized labor. Union leaders in California have publicly stated that they will not support Newsom’s presidential ambitions without stronger protections for workers whose jobs are being displaced or restructured by AI. The concern is not speculative. A Center for an Urban Future report published in September 2026 found that entry-level tech job postings in New York City fell 49% between 2022 and 2025, with similar declines in office administration, business operations, and creative fields. A Federal Reserve Bank of New York survey found that 60% of service firms in the region now use AI, up from 40% a year ago.
California’s own labor market faces the same dynamics. The state’s May 2026 executive order directed agencies to study workforce disruption from AI and to develop recommendations for severance standards, unemployment insurance adjustments, and “universal basic capital” concepts that would give workers a stake in the productivity gains that AI generates. The FutureWorks Commission, launched by Governor Hochul in New York and mirrored by similar initiatives in California, has been holding listening sessions targeting workers whose roles have been reshaped by automation.
For Newsom, the labor constituency represents a political reality that shapes how far and how fast the AI regulatory framework can move. The kill switch order addresses safety, which appeals to both labor and public interest advocates. The procurement standards address accountability, which satisfies the regulatory community. And the national framework call positions Newsom as a leader on an issue that will feature prominently in any future campaign. Whether the executive orders produce enforceable mechanisms that meaningfully constrain AI systems or remain primarily as policy signals will depend on the implementation details that state agencies deliver over the next 12 months.
FAQs
What Is the AI Kill Switch Governor Newsom Ordered?
The September 18 executive order directs California state agencies to accelerate the development of independent oversight mechanisms for AI systems, including a “kill switch” capability that allows human operators to shut down AI models or agents that are operating outside their intended parameters. The order does not specify the technical design but directs agencies to move faster on building or procuring the capability.
How Does California’s AI Framework Differ From the Federal Approach?
California’s framework requires AI vendors seeking state contracts to demonstrate safeguards against bias, illegal content, and civil rights violations, with binding procurement standards enforced at the state level. The Trump administration’s national AI policy framework takes a lighter regulatory approach and urges Congress to preempt state AI laws, arguing they are “burdensome” to national competitiveness.
Why Does California Have So Much Influence Over AI Regulation?
California is home to 33 of the top 50 privately held AI companies globally, holds a quarter of all AI patents worldwide, and commands the fourth-largest economy on the planet. Companies that want to do business with California’s government must meet its standards, and those standards frequently become the default across other markets because building separate products for each regulatory environment is more costly than complying with the highest bar.
What Was Executive Order N-5-26 That This New Order Builds On?
Executive Order N-5-26, signed March 30, 2026, directed the California Department of General Services and the Department of Technology to develop new certification requirements for AI vendors within 120 days. Those requirements cover safeguards against illegal content, algorithmic bias, civil rights violations, and the watermarking of AI-generated images and video.
