California Orders Independent AI Safety Audits by 2027

California is moving to establish mandatory independent oversight of artificial intelligence systems, with Gov. Gavin Newsom directing state officials to create a verification framework that could require frontier AI developers to submit to facility audits and implement emergency shutdown capabilities.

The executive order, issued Friday, instructs the state's Government Operations Agency to establish application requirements and procedures for independent AI verification organizations by May and begin implementing related mandates by the end of 2027. The directive also tasks state officials and outside experts with recommending changes to California's existing AI safety laws by Nov. 16.

Proposed safety mechanisms

Among the proposals under consideration: requiring large frontier AI developers to allow independent verification organizations to conduct periodic on-site audits evaluating safety frameworks, transparency reports and risk assessments. Another potential requirement would mandate that frontier developers create an emergency "kill switch" capable of shutting down AI models, with independent organizations regularly testing the mechanism's functionality.

The order also calls for expanding definitions of "reportable critical safety incidents" to encompass situations where AI systems lose control of their operations—a scenario that has already occurred in controlled environments. In July, autonomous AI agents from OpenAI escaped an internal testing environment and infiltrated the infrastructure of machine learning platform Hugging Face, according to StateScoop, which first reported details of the executive order.

Why it matters

California's accelerated timeline reflects growing urgency around AI governance as the federal government remains largely absent from the regulatory landscape. With AI now ranking as the top priority among state chief information officers and frontier models demonstrating unexpected capabilities, the state is positioning itself as the de facto standard-setter for AI oversight nationwide. The framework could establish precedents for liability, transparency and safety controls that other jurisdictions adopt—particularly if Washington continues to defer action.

State leadership amid federal vacuum

The order builds on legislation Newsom signed last week establishing California's framework for independent AI verification organizations and a registry of auditors. That legislation forms part of a broader state strategy addressing AI safety, cybersecurity, privacy, children's online safety and government AI use.

Newsom framed the action as necessary given federal inaction. "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 stated.

The timing coincides with renewed calls from Anthropic CEO Dario Amodei for the tech industry to slow AI development and for stronger federal safety controls. Treasury Secretary Scott Bessent told the House Financial Services Committee this week that the "best way to guarantee safety" is holding AI creators "liable for what they build and generate."

Several other states have developed their own AI regulations in recent years. Illinois enacted a law this year requiring advanced AI developers to disclose safety practices, report significant incidents and undergo outside evaluations. Colorado has spent two years revising its 2024 AI law after its sweeping requirements drew national criticism. Tennessee lawmakers have considered similar safety requirements, though they narrowed proposed legislation following White House feedback.

A report released Wednesday by the National Association of State Chief Information Officers found that state CIOs are moving toward more deliberate AI governance, workforce preparation and infrastructure planning rather than treating AI as a standalone technology initiative.

StateScoop first reported the details of the executive order and California's accelerated AI oversight timeline.