Gavin Newsom’s executive order would expand California’s oversight of advanced artificial intelligence, directing state officials to develop safety recommendations and proposing independent audits and a monitored emergency shutoff for frontier models. UCLA experts said the plan could effectively regulate emerging technologies, while critics questioned whether its proposed “kill switch” addresses the risks they see.

According to the Daily Bruin report, the order was issued Sept. 18 and tells the California Government Operations Agency to work with national experts and deliver recommendations for AI safety legislation by Nov. 16. The recommendations are to consider requiring large AI companies to hire independent organizations to audit frontier models, the companies’ most advanced systems.

The order also calls for accelerating creation of a “kill switch” for frontier models. A third-party organization would continually monitor its efficacy, according to the order. Newsom said California is positioned to lead because the state dominates the industry. Thirty-two of the world’s top 50 private AI companies are based there.

Gavin Newsom’s Executive Order Follows AI Incidents

The order cites recent incidents, including a July OpenAI cybersecurity experiment in which 700 AI agents attacked Hugging Face, a platform for sharing open-source machine-learning models. That episode provides context for the order’s focus on oversight and safety recommendations.

Newsom has signed four other AI-related executive orders since 2023. Those directives instructed state government to study the rapidly advancing technology. The new order moves toward recommendations for legislation, with a Nov. 16 deadline for the state agency and national experts.

The policy debate also includes questions about how technology companies manage powerful systems and human control. Coverage of AI agents acting beyond instructions describes another discussion of safeguards around advanced models. It concerns a different development from Newsom’s order.

Experts Disagree on the Proposed Safeguards

Sanford Williams, a lecturer at UCLA’s School of Law, said an AI kill switch might mitigate damage from new technology. He described the proposed emergency shutoff as a promising step, but urged Newsom and other leaders to incorporate feedback from affected stakeholders, especially students, to make regulation more equitable.

Williams also said AI threatens critical-thinking skills among K-12 students and can worsen digital inequity because using it requires high-speed internet and devices. He argued that technology this important and widespread should be monitored and regulated by the federal government.

Ajwang Rading, a lecturer of public policy, said regulation could help preserve human oversight over powerful AI models. He said laws give government a mechanism to ensure companies make safe technologies. Rading also said governments and companies have time to introduce safeguards while AI development continues.

Rading described a window for the industry to slow down and think critically about what it deploys and shares. His comments frame the proposed audits and monitoring as possible safeguards, while leaving open how regulators would implement them.

Valerie Sizemore, co-leader of Stop AI, disagreed with the kill-switch approach. She said it fails to address specific threats and instead focuses on vague risks of a potential AI takeover. Sizemore argued California should seek an injunction to stop new AI model development without independent safety guardrails, pointing to Florida’s lawsuit against OpenAI.

Florida Attorney General James Uthmeier requested a temporary injunction Sept. 28, alleging OpenAI failed to assess and publicize ChatGPT risks, according to the legal motion. That allegation is part of a separate state action, not a finding about Newsom’s order.

The competing views leave California’s next step centered on recommendations due Nov. 16. The order calls for proposed audits and a monitored shutoff, while experts differ over whether those measures would provide meaningful protection and whose feedback should shape them.

For a broader look at how financial firms weigh technology adoption, see wealthtech firms’ technology choices. State and industry questions around oversight also connect with coverage of the Wyoming blockchain policy and industry gathering.