California Pioneers Independent AI Oversight, Calling for National Adoption

California Governor Gavin Newsom has appointed four world-leading experts to guide the implementation of his groundbreaking AI executive order, which accelerates independent oversight and explores mandatory kill switches for frontier AI models.

Newsom announced today that Jason Goldman, Gillian Hadfield, Alondra Nelson, and Rob Reich will advise California’s work on some of the most challenging questions in AI governance. The appointments come as California positions itself as the nation’s standard-bearer for AI safety regulation, filling a vacuum left by federal inaction and building on the state’s already extensive framework of AI laws.

The four experts bring heavyweight credentials. Goldman served as the first White House Chief Digital Officer and helped build early products at Google and Twitter. Hadfield focuses on AI alignment and regulatory system design at Johns Hopkins University. Nelson previously led the White House Office of Science and Technology Policy and now directs the Science, Technology, and Social Values Lab at the Institute for Advanced Study. Reich, a Stanford professor, spent time as a senior advisor to the United States AI Safety Institute.

The expert panel will tackle questions that have divided the AI industry and policymakers: who verifies that frontier AI companies are telling the truth about their safety claims, and whether advanced models can be reliably shut down when needed. These aren’t theoretical concerns anymore. Recent incidents have alarmed Americans and industry leaders alike, prompting calls for stricter government regulation that the federal government has ignored.

Newsom’s executive order directs the Government Operations Agency to speed up implementation of recently passed state laws and work with the Governor’s Office of Emergency Services to develop recommendations within two months. Those recommendations could require frontier AI companies to embed independent verification organizations directly inside their labs to conduct regular audits and evaluations. That’s onsite oversight, not occasional check-ins.

The order also calls for verification of the safety frameworks, transparency reports, and risk assessments that frontier AI companies must file under state law. But here’s the critical part: those documents would need to meet standards deemed adequate by independent verification organizations, not the companies themselves. California wants third-party eyes on everything.

Then there’s the kill switch. The executive order explicitly advances the creation of emergency shutdown mechanisms for frontier models. More importantly, it requires that the efficacy of those switches be verified on an ongoing basis by independent organizations. A kill switch that doesn’t work isn’t a safety measure. It’s theater.

California Governor Gavin Newsom has appointed Jason Goldman, Gillian Hadfield, Alondra Nelson, and Rob Reich to help implement an executive order accelerating independent oversight of frontier AI. Their two-month assignment includes recommendations on onsite verification, safety-report audits, expanded definitions of critical incidents, and reliable emergency shutdown mechanisms. Building on California’s AI laws, the initiative positions the state as a potential national model while federal regulation remains stalled.

Newsom signed two bills in September 2026 that established the foundation for this work. Senate Bill 813 created the nation’s first framework for certifying independent verification organizations with enough expertise and demonstrated independence from AI companies to objectively assess AI systems and models for safety and risk. Assembly Bill 1405 established a state registry for AI auditors that assess AI systems for compliance with state law, including standards for their independence, transparency, and integrity.

Those laws were already first-in-the-nation measures. The executive order takes them further by accelerating their implementation timelines and pushing for stronger requirements. The order also directs experts to update definitions of critical safety incidents to include loss-of-control incidents like the recent Hugging Face attack.

California’s approach reflects lessons learned from the state’s long history of technology regulation. In 2024, Newsom vetoed a bill that would have imposed stricter pre-release safety testing and kill switch requirements on certain large AI models. Critics argued that bill went too far. The current framework, built through SB 53 in 2025 and strengthened in 2026, emphasizes transparency and accountability without freezing innovation.

Newsom’s executive order arrives at a moment when federal AI policy has stalled. No federal law requires AI companies to report dangerous incidents when they happen. President Trump has rejected calls for new regulations that would increase security and protect public safety. The administration’s contribution to AI policy has been to rename artificial intelligence as “super intelligence.”

That vacuum leaves states to act on their own. California has moved aggressively to fill the gap. The state now has laws addressing nearly every major area of AI policy: frontier-model safety, independent oversight, children and companion chatbots, privacy, deepfakes, fraud, cybersecurity, workers, and government deployment.

The contrast with federal inaction is stark. Goldman put it plainly: “Governor Newsom’s executive order addresses the hardest questions in AI safety governance, including who verifies a frontier lab’s safety claims and whether a model can be reliably shut down. California continues to show how democratic oversight can apply to one of the world’s most consequential industries.”

Hadfield emphasized the urgency. “The need for effective, accountable, and independent oversight of increasingly powerful and autonomous AI is urgent and I’m very glad to join this great team to quickly help ensure we steer in the right direction.”

California’s most consequential move is not the phrase “kill switch,” but the decision to make independent verification a continuing institutional requirement. Safety claims, transparency reports, and shutdown mechanisms are only meaningful when outsiders can test them repeatedly and challenge the companies responsible for building the systems. If California can turn that principle into workable standards without creating an impossible compliance burden, it may establish a durable template for governing frontier AI.

California’s AI regulatory framework didn’t appear overnight. It’s been built piece by piece since Newsom issued the nation’s first executive order on generative AI in 2023. The state’s current approach covers terrain that other states haven’t even started mapping.

On frontier AI safety, California enacted SB 53 in 2025, requiring frontier AI developers to publicly disclose safety frameworks, report specified critical safety incidents to the state, and protect whistleblowers who report serious risks. That was the nation’s first AI safety law.

On child safety, California passed the bipartisan Adam’s Law this year, requiring safeguards around companion chatbots used by children. The law mandates suicide and self-harm crisis protocols, parental controls, and notifications when safety settings are disabled. California also banned toys containing AI companion chatbots for five years and prohibited using student personal information to train AI models.

On social media and children, California enacted the strongest-in-nation legislation prohibiting social-media companies from providing users under 16 with addictive features including autoplay and algorithmic feeds based on user history and profiles. The state requires age verification signals and prohibits platforms from knowingly providing addictive feeds to minors without parental consent.

On deepfakes and synthetic content, California banned sexually explicit deepfakes, required AI watermarking, addressed deceptive AI-generated election content, and required transparency around AI training data.

On privacy, California passed the Delete Act, allowing Californians to direct registered data brokers to delete their personal information. The state launched DROP, which lets people submit a single deletion request to registered data brokers. Newsom also issued an executive order in March 2026 strengthening privacy and civil-liberties protections for state AI procurement.

On workers and AI disruption, California issued an executive order addressing AI’s economic and labor impacts, including expanded tracking of AI’s economic effects and work to adapt job-training and public-benefit programs. The state launched an AI Unemployment Tracker and solicited statewide public input through Engaged California.

On cybersecurity, California launched the nation’s first AI Cyber Defense Program to use AI to identify vulnerabilities, harden networks, and support incident response. The state is establishing an AI Cybersecurity Officer in each agency and working with frontier AI companies and critical-infrastructure partners on response protocols.

On AI-enabled fraud, California launched a Tech Fraud Task Force to combat AI-enabled fraud and addressed algorithmic pricing through state antitrust law.

The experts joining Newsom’s advisory group understand what’s at stake. Nelson framed the challenge clearly: “California has long led on technology governance, and this executive action asks exactly the right questions: not whether to govern AI, but how. Independent verification, AI auditing, and mandatory transparency are the kinds of durable institutional mechanisms that can govern rapid technological change and hold powerful actors accountable, so their tools serve humanity instead of endangering it.”

Reich connected safety to innovation: “Innovation and safety go hand in hand in any responsible marketplace, and AI is no different. In order to gain the enormous potential benefits from AI, we need governance practices that establish trust in the technology. I am honored to join the effort in California to accelerate the scientific practice of AI testing and evaluation and establish robust frameworks of independent assessment and oversight of AI systems.”

These perspectives matter because they come from people who have worked inside government and academia, who understand both the promise of AI and the risks of letting powerful technology develop without accountability. They’re not calling for AI to be banned or frozen. They’re calling for verification, transparency, and reliable safety mechanisms.

Newsom is calling on Congress and President Trump to review and adopt California’s framework, or use its nation-leading regulation as a floor for federal policy that would benefit all Americans. That’s a direct challenge to Washington’s inaction.

California’s approach could serve as a template. The state has shown that comprehensive AI governance is possible without crushing innovation. More than 20 AI laws went into effect in California in 2026, covering everything from pricing algorithms to data transparency. Enforcement involves the California Privacy Protection Agency and the Attorney General, creating multiple pathways for accountability.

The question isn’t whether California’s model will influence national policy. The question is when. States have often led on technology regulation when the federal government wouldn’t act. California’s regulations on vehicle emissions, data privacy, and consumer protection have repeatedly shaped national standards. AI regulation could follow the same path.

Other states are watching. California’s legislative approach balances promoting innovation and addressing AI risks, serving as a state-level experiment that provides real data on what works. The emphasis on independent verification organizations and third-party audits creates accountability without requiring government agencies to become AI experts themselves.

The expert panel has two months to develop recommendations for strengthening state law. Those recommendations will likely push California even further ahead of other states and the federal government on AI oversight. The emphasis on embedding independent verification organizations directly inside frontier AI labs represents a significant escalation in oversight intensity.

Requiring ongoing verification of kill switch efficacy addresses a problem that has plagued other safety mechanisms: they work until they don’t. Continuous verification means that companies can’t just flip a switch during an initial inspection and then let the mechanism degrade over time.

Updating definitions of critical safety incidents to include loss-of-control incidents recognizes that AI risks have evolved. The Hugging Face attack demonstrated that AI systems can be compromised in ways that weren’t fully anticipated when earlier safety frameworks were written. California’s willingness to update its definitions shows an understanding that AI governance needs to adapt as quickly as the technology does.

California’s leadership creates pressure on other states and the federal government to act. Tech companies operating nationally can’t ignore regulations in a state that hosts much of the AI industry. What California requires becomes a de facto standard, particularly when the federal government won’t establish one.

The appointments announced today signal that California is serious about implementation. Goldman, Hadfield, Nelson, and Reich aren’t symbolic choices. They’re experts with deep knowledge of AI governance, regulatory design, and the intersection of technology and policy. Their recommendations will carry weight because they understand both the technical challenges and the institutional mechanisms needed to address them.

California has established a comprehensive framework for AI governance that other states and the federal government should adopt. The state’s emphasis on independent oversight, mandatory transparency, and verified safety mechanisms creates accountability for one of the most consequential technologies of our time. With federal action stalled, California’s leadership provides a model for democratic oversight of AI that puts public safety ahead of industry convenience.