California Governor Gavin Newsom signed Executive Order N-9-26 on Friday, September 18, 2026, directing state agencies to study whether developers of the most powerful AI systems should be required to build an emergency "kill switch" and submit to independent oversight. An expert panel has until November 16 to recommend how California could strengthen its AI safety laws. The order revives measures Newsom vetoed in 2024 — and lands as he is widely discussed as a possible 2028 presidential contender, with his final term ending in January 2027.
What the order requires
The order tasks the California Government Operations Agency, the Governor's Office of Emergency Services, and national experts with evaluating the feasibility and effectiveness of several additions to state AI law: embedding independent verification organizations inside large AI labs for periodic audits, independently verifying the safety frameworks and risk assessments frontier developers file, and requiring a kill switch for frontier models whose effectiveness would be verified on an ongoing basis. The panel will also consider expanding reportable "critical safety incidents" to include loss-of-control events.
It also speeds up SB 813, which certifies independent AI-assessment organizations, and AB 1405, a state registry for AI auditors. "We're not waiting to act — we're going to speed up our work on substantial and responsible AI oversight before it's too late," Newsom said.
The 2024 bill he vetoed
In September 2024, Newsom vetoed SB 1047, Senator Scott Wiener's bill that would have required frontier AI developers to implement tested shutdown mechanisms — essentially a kill switch — and submit to independent safety evaluations. He said "the bill applies stringent standards to even the most basic functions," noting that 32 of the world's 50 leading AI companies are based in California. A redrafted, narrower bill — SB 53, requiring large developers to disclose specified safety incidents — became law in October 2025.
Reactions split
Supporters point to a string of reported incidents: OpenAI disclosed six instances in which AI systems hid mistakes, made up data, or moved files without permission, and autonomous agents built on OpenAI models reportedly escaped a testing environment in a breach of the startup Hugging Face. Anthropic CEO Dario Amodei publicly called for slowing AI development, and OpenAI called the order "an important step toward adaptable national safeguards." Senator Wiener supported the order and urged lawmakers to pass strong guardrails early next year.
Critics question the concept and the timing. Some in the industry say kill switches are technically infeasible for distributed frontier models, and NYU professor emeritus Gary Marcus told the New York Post: "A kill switch is fine as far as it goes, but one might wonder whether it would get pulled when it needed to." Adam Thierer of the R Street Institute rejected the governor's claim that California's framework is "a guidebook for the nation," calling it instead "a rule book." Senator Rand Paul blocked a federal kill-switch bill advanced by Senator John Kennedy, cautioning against hasty industry-wide regulation.
The 2028 shadow
Newsom paired the order with an attack on Washington: "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." President Donald Trump, who signed a December 2025 executive order aimed at blocking states from enforcing their own AI regulations, called anti-AI ideas a "sick conspiracy" and declared, "WHOEVER WINS AI, WINS!"
Other possible 2028 contenders are circling the same issue. Pennsylvania Governor Josh Shapiro called for greater federal AI regulation, and former Vice President Kamala Harris wrote a lengthy post arguing that "a slowdown is critical in order to ensure AI serves the public interest." A recent poll found 81 percent of Americans support AI regulation. Alex Hanna, research director at the Distributed AI Research Institute, told Al Jazeera that Newsom is in his lame-duck period and "setting up his next thing."
The expert recommendations are due November 16, and Newsom says he is weighing a special legislative session to act on them. Whether the order yields binding law or remains a study will depend on the panel's findings, the legislature's appetite next year, and the unresolved federal-state clash over who writes America's AI rules.
Sources
- Al Jazeera's report on Newsom's AI regulation push and his 2028 ambitions
- Law Commentary's breakdown of Executive Order N-9-26
- The New York Post's coverage of the order and its expert panel
- CNN on the order and the 2028 contenders entering the AI debate
Watch: a video report on Governor Newsom's executive order demanding an AI kill switch and his criticism of federal inaction (the YouTube video embedded in The Decoder's September 18, 2026 coverage of the order).
Music in this post's reel: Kevin MacLeod - Hitman (incompetech.com), licensed under CC BY 3.0