TL;DR
California Governor Gavin Newsom has named a four-person expert panel to operationalize his recent AI executive order, including developing a "kill switch" for rogue frontier AI models and embedding independent third-party monitors inside AI companies.
What happened
- Newsom appointed four advisors on September 23, 2026, to guide implementation of his AI executive order issued days earlier.
- Panel members: Jason Goldman (first White House Chief Digital Officer, Google/Twitter alum), Gillian Hadfield (Johns Hopkins, AI alignment), Alondra Nelson (Institute for Advanced Study, former acting White House OSTP director), and Rob Reich (Stanford, former US AI Safety Institute senior advisor).
- The order directs the panel to advance two concrete mechanisms: mandatory emergency shutoff ("kill switch") capabilities for frontier models and independent third-party embeds inside frontier AI companies to verify safety claims.
- California has already enacted supporting legislation: SB 813 creates a framework for independent verification organizations, and AB 1405 establishes a state registry for AI auditors.
Why it matters
- California is filling the federal vacuum: Newsom explicitly frames the action as a response to the Trump administration's inaction, positioning the state as the de facto national AI safety regulator.
- A legally mandated kill switch for frontier models would be a world first, forcing labs like OpenAI, Anthropic, and Google DeepMind (all California-based) to architect controllable shutdown paths into production systems.
- Independent third-party embeds inside AI companies would be a structural shift in oversight, moving beyond self-reported safety evaluations to externally verified ones.
- The panel's combined credentials span AI alignment research, legal system design, science policy, and democratic governance, giving recommendations cross-disciplinary credibility that is harder to dismiss.
- SB 813 and AB 1405 already on the books mean the panel is building on enacted law, not just advisory proposals, raising the probability of binding outcomes.
What to watch next
- Panel's first recommendations: The executive order sets a timeline; watch for interim reports specifying technical standards for kill-switch architecture and criteria for qualifying independent verification organizations.
- Industry response from frontier labs: Whether OpenAI, Anthropic, Google, and Meta publicly engage or legally challenge the embedded-monitor requirement will signal how hard this fight becomes.
- Federal preemption risk: Any move by Congress or the Trump administration to preempt California AI regulation would directly threaten this framework's enforceability.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.