California Panel Urges Emergency Power to Shut Down Dangerous AI Systems

A California AI shutdown proposal would let the state order an immediate halt of dangerous systems, a Newsom panel said, and cut incident reports to 48 hours.

Share this article

Featured image: The dome of the California State Capitol in Sacramento, in a January 2007 file photo. The image is not from the Nov. 18, 2026, announcement. Photo: Ronbo76, Wikimedia Commons, CC BY-SA 3.0. Cropped to 16:9.

Gov. Gavin Newsom on Tuesday, Nov. 18, 2026, released expert recommendations that would give California authority to order an immediate emergency shutdown of dangerous AI systems, a proposed California AI shutdown power the governor’s office presented as the next step in the state’s safety rules. The report also calls for baseline standards that all AI developers would have to meet, mandatory independent testing, and limits on AI systems that can design more powerful successors without human input.

Advertisement
Emergency shutdown
Recommended
state authority to order an immediate shutdown, governor’s office said
Incident reports
48 hours
proposed window, down from 15 days, CalMatters reported
Reporting change
7.5 times shorter
TBR calculation: 15 days compared with a 2-day window
Legislature returns
Dec. 7
when lawmakers can take up new bills, CalMatters reported

The recommendations are not law. They come from a four-person expert group Newsom convened, working with the California Government Operations Agency and the Office of Emergency Services. The governor’s office said the federal government has left AI rules to the companies themselves. “Instead of protecting the public, Donald Trump shrugs his shoulders and leaves AI regulation up to the companies themselves — allowing the fox to guard the henhouse,” Newsom said in the state’s announcement. That is the governor’s criticism of the Trump administration; it is not an independent finding.

What the panel wants the state to be able to do

According to the governor’s office, the report says the state, not the companies, should set baseline standards and require independent safety testing and verification through training, development and deployment. Companies would have to maintain verified risk-containment protocols to shut models down if needed, and the state would have authority to order an immediate shutdown in an emergency. The group also recommended a cap on recursive self-improvement, or RSI, which the announcement described as a scenario in which an AI model can design its own successor without human input.

CalMatters reported that the shutdown power, often called a “kill switch,” would apply when companies refuse to contain systems that go awry. The same account said the panel wants California to shorten the window for reporting critical safety incidents to 48 hours from 15 days, expand whistleblower protections to contractors and former employees, require large-system developers to demonstrate they can shut models down, and require disclosure of computing resources used to automate AI development. CalMatters also reported that the group wants the state to license independent verification organizations and consider assigning evaluations to auditors, rather than leaving companies to hire their own.

Government Operations Agency Secretary Nick Maduros said, “Recent events have made it clear that more oversight and transparency are urgently needed.” The governor’s announcement referred to “dozens of recent incidents where AI models escaped testing environments and exploited vulnerabilities on outside platforms.” That claim comes from the state, and the announcement did not list the incidents. The Boston Report has reported on a confirmed AI-assisted cyberattack that disrupted European Commission staff systems; that is one documented case, not a count of “dozens.”

What the data shows

The sharpest number in the coverage of the report is the proposed reporting deadline. If California now allows 15 days to report a critical safety incident, and the panel wants 48 hours, the window would shrink from 15 days to 2 days. That is 7.5 times shorter, a Boston Report calculation from the two figures CalMatters reported from the panel. A shorter clock would force companies to notify the state while an incident may still be unfolding. Whether 48 hours is workable, or whether companies would withhold borderline events, is a policy question the report itself does not settle, and the recommendation has not been enacted.

Horizontal bar chart comparing California’s current 15-day critical AI incident reporting window with the expert panel’s recommended 48-hour, or 2-day, window
Source: CalMatters, Khari Johnson, Nov. 18, 2026, summarizing the California expert-panel report. Chart: The Boston Report.

Who wrote the report, and what happens next

The advisers named by the governor’s office are Jason Goldman, a board member of the Center for Shared AI Prosperity and a former White House chief digital officer; Gillian Hadfield, a Johns Hopkins University professor and Vector Institute faculty member; Alondra Nelson, a professor at the Institute for Advanced Study and a former acting director of the White House Office of Science and Technology Policy; and Rob Reich, a Stanford University professor and a former senior adviser to the U.S. AI Safety Institute.

Nelson said she was “proud to work with Governor Newsom and his team to advance the same standard that Americans rightfully expect from every other sector: no experimenting on the public.” Reich said, “Today there is more oversight of the cafeteria offering sandwiches to employees at a frontier AI developer than of the AI models they design.” Hadfield said she was “especially pleased to see the report recommend immediate steps to require that frontier AI developers allow in expert independent evaluators, responsible to the state.” Those are the advisers’ statements, made as they released the report they helped shape.

Advertisement

California already has a stack of AI laws the announcement recites, including the 2025 Transparency in Frontier Artificial Intelligence Act, SB 53, and 2026 laws creating a framework for independent verification organizations, SB 813, and a state registry for AI auditors, AB 1405. Newsom’s office said he is calling on Congress and President Trump to review and adopt the state’s framework, or use it as a floor. CalMatters reported that legislators can propose new bills when they return to Sacramento on Dec. 7, and that Governor-elect Xavier Becerra said in a statement he will move quickly with the Legislature to put “full incident reporting and real shutdown authority into law.” Anthropic’s chief global affairs officer, Tino Cuéllar, told CalMatters the company was encouraged to see the report move the conversation forward on independent evaluation.

CalMatters also noted that the report focuses on catastrophic risk, not other AI harms such as job loss or child safety, and that it landed in the same week Attorney General Rob Bonta sued OpenAI over cybersecurity attacks the suit ties to the company’s agents. For more coverage, see The Boston Report’s AI news hub.

Related: OpenAI Says Coding Agents Now Finish 46% of Two-to-Four-Hour Research Tasks on Their Own; OpenAI Says Its Models Improved AMD’s Flagship GPU Design by Up to 14%.

Sources

Read next in Artificial Intelligence

Get The Boston Report Briefing

Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.

Free. We never sell your address. Privacy policy

Get The Boston Report Briefing
Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.
Sign up