TL;DR
California Governor Gavin Newsom signed an executive order on September 18, 2026 directing the state to accelerate independent AI oversight and advance a mandatory "kill switch" for frontier models, explicitly positioning California's framework as the national standard Washington has refused to create.
What happened
- Newsom signed an executive order directing the Government Operations Agency to accelerate implementation of two recently passed laws: SB 813 (independent verification organizations) and AB 1405 (state registry for AI auditors).
- Expert panel convened with a two-month deadline to recommend how to strengthen state law, including requiring onsite independent auditors embedded inside frontier AI labs.
- "Kill switch" mandate proposed requiring frontier AI companies to develop emergency shutoff mechanisms, with ongoing efficacy verified by independent third parties.
- Hugging Face attack cited as a triggering incident, with the order calling for updated definitions of critical safety incidents to include loss-of-control events.
- Federal inaction named directly: no federal law currently requires AI companies to report dangerous incidents, and the Trump administration has rejected new AI safety regulations.
Why it matters
- First-in-nation kill switch requirement, if enacted, would force frontier AI developers operating in California to build and maintain verifiable emergency shutoffs, a structural constraint no other jurisdiction has imposed.
- Onsite independent auditors embedded in AI labs would represent a step-change from self-reported safety frameworks, directly targeting the trust gap between companies and regulators.
- California's regulatory surface is already broad: laws now cover frontier-model safety, independent oversight, children and companion chatbots, privacy, deepfakes, fraud, cybersecurity, workers, and government deployment, making it the de facto national baseline by market weight alone.
- Newsom is explicitly lobbying Congress to adopt California's framework as a federal floor, escalating a state-versus-federal tension that could reshape how frontier AI companies structure compliance nationwide.
- AI CEOs calling for regulation while Washington does nothing creates political cover for California to move faster, reducing industry resistance that typically slows state-level tech regulation.
What to watch next
- Expert panel recommendations due within two months: whether the kill switch and onsite auditor proposals survive into binding legislation will determine if this order has teeth or is political positioning.
- Frontier lab responses: companies like OpenAI, Anthropic, and Google DeepMind, all California-headquartered, will face the most direct compliance pressure and may signal relocation or legal challenges.
- Federal preemption risk: if Congress acts on AI regulation before California's framework is fully codified, a weaker federal standard could preempt the stronger state rules, the exact scenario Newsom is trying to prevent by framing California as the floor.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.