Full Breakdown
California Governor Accelerates AI Oversight and Pursues “Kill Switch” Initiative
By Drooid · · How we work
Core Event
In September 2026 Governor Gavin Newsom signed an executive order that fast-tracks two AI-related laws passed earlier this year and creates a two-month mandate for a panel of “world-leading experts” to draft recommendations on strengthening state AI safety rules. The order calls for exploring a “kill switch” that could shut down frontier AI models that behave unpredictably, and it moves up the implementation timeline for Senate Bill 813 and Assembly Bill 1405, which establish independent AI auditors and a state-wide auditor registry.
Background & Context
California has built a “nation-leading” AI regulatory framework since a 2023 executive order on generative AI. In 2024 Newsom vetoed a bill that would have required pre-release safety testing and a kill switch for large models. The following year the state enacted SB 53, the Transparency in Frontier Artificial Intelligence Act, mandating disclosure of safety frameworks and reporting of critical incidents. SB 53 was complemented by SB 813, creating a certification system for independent verification organizations, and AB 1405, establishing a public registry for AI auditors.
Recent safety alerts, including a “Hugging Face attack” that demonstrated loss-of-control behavior in an OpenAI model, and a resignation from Anthropic in which researcher Jacob Coxon warned that AI developers were “gambling with our lives,” have intensified calls for stronger guardrails.
Timeline
- June 4, 2026 – Newsom attends the 94th Annual Meeting of the United States Conference of Mayors in Long Beach, California.
- 2023 – First executive order on generative AI.
- 2024 – Veto of a bill requiring a kill switch for certain AI models.
- 2025 – Passage of SB 53, the nation’s first frontier-AI safety law.
- Early 2026 – Enactment of SB 813 and AB 1405, establishing independent AI audit mechanisms.
- September 2026 – Executive order issued to accelerate oversight and explore a kill switch.
Official Statements & Responses
Trump has publicly dismissed AI safety warnings as a “hoax,” and congressional action on AI regulation remains stalled ahead of the 2026 midterm elections.
Criticism & Opposition
Anthropic CEO Dario Amodei cautioned that a kill switch “could be a good idea” but is no “panacea,” reflecting industry skepticism about technical feasibility. Several smaller AI firms have warned that mandatory shutdown mechanisms could hinder competition and innovation.
Conflicting Reports & Gaps
Experts differ on whether a functional kill switch is technically achievable. While Newsom’s order treats the concept as a policy goal, Amodei and other technologists argue that current technology does not support a reliable, universal shutdown for frontier models. The order also leaves open how independent verification organizations will be selected and what standards will define “efficacy” of a kill switch.
Why It Matters / Impact
California’s accelerated timeline positions the state as a potential template for federal AI regulation, especially as many leading AI firms are headquartered there. If the kill-switch framework gains traction, it could set de-facto standards for global AI developers, influencing safety practices and competitive dynamics.
What’s Next
The expert panel is tasked with delivering a guide within two months. The state’s accelerated implementation of SB 813 and AB 1405 will begin in December 2027, ahead of the original January 2029 schedule.