TL;DR
California Governor Newsom signed an executive order accelerating two new AI laws and directing agencies to explore mandatory kill-switch mechanisms for frontier AI models, putting the world's fifth-largest economy at the center of AI safety regulation.
What happened
- Newsom issued an executive order on September 18, 2026 targeting frontier AI governance in California.
- The order accelerates implementation of two recently signed AI laws, moving faster than their standard legislative timelines.
- Agencies are directed to explore mandatory kill-switch mechanisms that could force developers to shut down or constrain frontier AI models on demand.
- The order applies to frontier AI models, the most capable and potentially highest-risk systems, not narrow or legacy software.
Why it matters
- California is the home jurisdiction for most major AI developers including Google, Anthropic, and Meta, meaning state rules carry outsized global weight.
- A mandatory kill switch would be the most interventionist AI safety requirement yet proposed by any major U.S. jurisdiction, going beyond disclosure or audit mandates.
- The order signals Newsom is shifting from AI-friendly to AI-cautious governance, a reversal from his 2024 veto of SB 1047, the earlier landmark safety bill.
- If enacted as regulation, kill-switch requirements could force architectural changes in how frontier models are deployed and updated, raising compliance costs.
- Other states and the EU will watch closely: California rules often become de facto national standards when companies choose uniform compliance over state-by-state fragmentation.
What to watch next
- Agency rulemaking timelines: watch for draft regulations translating the kill-switch concept into technical and legal specifics, which will reveal how broad or narrow the requirement actually is.
- Industry legal challenges: major AI developers may contest the order in court, and any injunction would signal the limits of state authority over AI infrastructure.
- Federal preemption dynamics: a federal AI framework moving through Congress could override California rules, making the next 12 months a race between state and federal action.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.