In a bold move that sidesteps a largely dormant federal response, Governor Gavin Newsom issued an executive order on September 18, 2026 to convene a panel of world-leading experts. Their charge: draft a practical guide that strengthens California’s AI safety regime, including the creation of an “AI kill-switch” capable of shutting down rogue frontier models. The order accelerates timelines for two landmark statutes—SB 813 and AB 1405—aimed at establishing independent verification bodies that can audit AI systems for risk and compliance.
The newly formed advisory group must deliver its recommendations within 60 days. Among the topics on the table are mandatory safety plans authored by third-party auditors, on-site verification organizations embedded in AI labs, and continuous testing of any emergency shut-off mechanism. The panel will also revisit the definition of a “critical safety incident” to explicitly cover loss-of-control events like the recent Hugging Face breach.
Legislative foundation that makes California a national model
California’s AI framework is built on a cascade of first-in-the-nation laws. In 2025, the state enacted SB 53, the Transparency in Frontier artificial intelligence Act, which forces developers of high-risk models to disclose safety architectures, report defined incidents, and protect whistleblowers. The following year, SB 813 created a certification process for independent verification organizations (IVOs) with the expertise and autonomy to evaluate AI products. Complementary AB 1405 established a public registry of AI auditors, setting standards for transparency, independence, and integrity.
Newsom’s current order pushes these statutes forward, mandating that frontier-AI firms host an IVO on their premises and that all safety documentation be validated by such an organization. It also requires the ongoing verification of any kill-switch technology, ensuring that the mechanism remains functional as models evolve. By tightening the regulatory feedback loop, the governor hopes to pre-empt the kind of unchecked escalation that industry insiders have warned could threaten humanity.
Federal inertia and the push for a national baseline
While California tightens its grip, the federal government remains without a comprehensive AI oversight law. No national statute compels companies to report dangerous incidents, and recent statements from the White House have offered little in the way of concrete action. This vacuum has spurred CEOs of major AI firms—OpenAI, Anthropic, and others—to publicly call for stricter regulation, yet Congress has stalled.
Newsom explicitly frames California’s approach as a template for the nation. In the order’s preamble, he argues that the state’s framework should serve as a “floor, not a ceiling” for federal policy, urging Congress and the president to adopt or at least reference California’s standards. By doing so, the governor hopes to protect Americans from the systemic risks posed by powerful models while preserving the innovative ecosystem that fuels Silicon Valley.
Political theater meets policy ambition
Beyond the regulatory agenda, the executive order arrives amid speculation about Newsom’s potential 2028 presidential bid. Recent media moments—such as a televised fly-fishing interview with CNN’s Jake Tapper—have painted the governor as both a rugged, relatable figure and a polished politician navigating the national spotlight. In that interview, Newsom hinted that a simultaneous candidacy from former Vice President Kamala Harris could trigger “mutually assured destruction” for both, underscoring the strategic calculations behind his policy moves.
Critics argue that the AI initiative doubles as political theater, showcasing Newsom’s capacity to act where the federal government cannot. Supporters, however, point to his track record of enacting concrete legislation—SB 53, SB 813, AB 1405—and now expediting their implementation. Whether the kill-switch proposal will survive legal challenges or industry pushback remains uncertain, but the governor’s determination to act swiftly reflects the high stakes he repeatedly cites.
As California charts a path toward mandatory independent audits and a functional AI shut-off, the state positions itself as the de-facto standard-setter for artificial intelligence safety in the United States. If federal lawmakers eventually follow suit, the nation could benefit from a unified safety net; if not, California’s framework may become the default rulebook for any AI firm operating on the West Coast.
