California signs Executive Order N-9-26 on AI verification and kill switches

On September 18, 2026, California Governor Gavin Newsom signed Executive Order N-9-26, directing the Government Operations Agency to dramatically accelerate the implementation of new third-party oversight of safety and security risks in AI systems, as well as independent audits.

The order also convenes a group of world-leading experts to provide, within two months, a guide for California to reinforce and strengthen its AI safety and security laws, with one proposal calling for the creation of a kill switch for frontier models, with the efficacy of the switch verified on an ongoing basis by an independent verification organization.

What it is

The order builds on SB 813 and AB 1405, which Newsom signed on September 9, 2026, creating a first-in-the-nation framework for independent AI verification organizations and a state registry for AI auditors.

On September 23, 2026, Newsom named Jason Goldman, Gillian Hadfield, Alondra Nelson and Rob Reich as the expert advisors shaping the recommendations, describing the panel as a group of experts at the intersection of technology, policy, and governance.

What it requires

The order directs officials to recommend potential amendments to California's AI safety and security laws, with that work due to the Government Operations Agency by November 16, 2026.

Four proposals are under study: embedding a designated independent verification organization onsite in frontier labs to conduct regular audits and evaluations.

A second proposal would require that the AI safety frameworks, transparency reports, and risk assessments that frontier AI companies are required to file under state law are verified pursuant to standards deemed adequate by an independent verification organization.

A fourth proposal would update definitions of critical safety incidents to include loss-of-control incidents such as the Hugging Face attack.

The order also moves up existing deadlines: the agency must complete and publicly post the application requirements, procedures, and criteria for independent verification organizations by May 1, 2027.

A related deadline directs the GOA to act on certain sections of Section 11549.82 of the Government Code by December 1, 2027 instead of the previously designated January 1, 2029 deadline, which concerns AB 1405's AI-auditor registry.

The order does not immediately impose those four requirements on private companies.

What security leaders should do

Frontier model developers should begin organizing the testing records, incident histories and control-ownership documentation that an onsite independent verification organization would likely request, since one proposal under study would embed such a reviewer inside the lab. Security teams should also map who would hold authority to trigger an emergency shutoff and how such a mechanism could be tested without disrupting production systems, given the order's focus on an independently verified kill switch. Because officials are weighing a broader definition of loss-of-control incidents, security leaders should review whether their own incident-classification criteria would currently capture an AI agent acting outside its intended scope. Legal and compliance teams should track the November 16, 2026 recommendations and the 2027 deadlines for SB 813 and AB 1405, since any resulting legislation could convert the current proposals into binding requirements.