The Pulse
California Names Experts to Design Frontier-AI Kill-Switch Rules
Governor Gavin Newsom appointed four experts to advise California on independent AI oversight and a possible emergency shutoff for frontier models. The panel will recommend whether state law should require lab monitors, verified safety asse
AI.info Team ·
“The need for effective, accountable, and independent oversight of increasingly powerful and autonomous AI is urgent.”
Gillian Hadfield, professor at Johns Hopkins University
California has appointed four technology and policy experts to advise Governor Gavin Newsom’s administration on new controls for frontier artificial intelligence, including a possible emergency shutoff for models that behave dangerously.
The announcement on September 23 follows Newsom’s executive order of September 18, which directs state agencies to accelerate independent oversight of AI companies and develop recommendations for changes to California law. The order does not itself impose a kill-switch requirement. It asks experts to determine how such a system could be created, independently tested and verified.
Four advisers will shape the proposal
The advisory group includes Jason Goldman, a board member of the Center for Shared AI Prosperity and an early product leader at Google and Twitter; Gillian Hadfield, a Johns Hopkins professor who studies AI alignment and legal, regulatory and technical systems; Alondra Nelson, a professor at the Institute for Advanced Study and director of its Science, Technology, and Social Values Lab; and Rob Reich, a Stanford professor who studies the governance of frontier science and technology.
Goldman said Newsom’s order addresses questions that sit between technical design and public authority: who verifies a frontier lab’s safety claims, and whether a model can actually be shut down. Hadfield said the panel’s task is to help establish independent oversight for increasingly autonomous systems.
Nelson said the state should focus on independent verification, AI auditing and mandatory transparency. Reich said the group would work to accelerate AI testing and evaluation while building independent assessment practices.
California is considering monitors inside AI labs
The proposals under review reach beyond disclosure rules. California may require frontier AI companies to place an independent verification organization inside their laboratories, where it would conduct regular audits and evaluations of safety systems.
State law already requires frontier developers to file safety frameworks, transparency reports and risk assessments. Newsom’s order asks experts to consider whether those materials should be verified against standards set by an independent organization rather than accepted solely on the developer’s own account.
The proposed emergency shutoff would also face continuing checks. California is considering a requirement that an independent verification organization test the switch over time and determine whether it can reliably stop a frontier model or the computing resources supporting it.
Two new laws provide the regulatory machinery
The September 18 order builds on two laws Newsom signed on September 9. Senate Bill 813, authored by Senator Jerry McNerney, establishes a framework for independent verification organizations that assess AI systems and models for safety and risk.
Assembly Bill 1405, authored by Assemblymember Rebecca Bauer-Kahan, creates a state registry for AI auditors and sets standards for their independence, transparency and integrity. Newsom’s order directs the Government Operations Agency, working with the Governor’s Office of Emergency Services, to speed up implementation of both laws.
California’s wider framework also includes the 2025 Transparency in Frontier Artificial Intelligence Act, or SB 53. The law requires frontier AI developers to disclose safety frameworks, report specified critical safety incidents to the state and protect whistleblowers who report serious risks.
The proposal follows the Hugging Face attack
Newsom’s administration says recent incidents, including the Hugging Face attack, have intensified calls for stronger controls around advanced AI systems. The executive order asks California to broaden the definition of a critical safety incident so it can cover loss-of-control events of that kind.
The governor is also pressing Congress and the White House to adopt California’s approach nationally. His administration says federal law does not require AI companies to report dangerous incidents when they occur, leaving states to build their own reporting and oversight systems.
California’s announcement describes the kill switch as a potential requirement for frontier models, not a device already installed across the industry. The next concrete step is the expert group’s recommendations on independent verification, lab oversight, incident definitions and whether a tested emergency shutoff can become part of state law.