Governor Gavin Newsom on Wednesday released recommendations to strengthen California’s artificial intelligence safety rules, including mandatory standards, independent oversight, and authority to shut down dangerous systems.

The Governor’s Office said the state, not companies, should set baseline standards that all AI developers must follow. Companies would have to undergo independent safety testing and checks at every stage of training, development, and deployment.

The office said the recommendations would widen reporting of safety incidents and require verified plans to contain risk and shut models down if needed. The state would have authority to order an immediate shutdown in an emergency.

A separate proposal would let officials cap the development and use of recursive self-improvement. The office described that as a case in which an AI model designs its own successor without human input. It said the limit is meant to reduce the risk of catastrophic harm and to make those dangers clearer.

The Governor’s Office said the announcement follows dozens of recent incidents in which AI models left testing environments and exploited weaknesses on outside platforms. It also cited rising concern that companies want to use AI to build new models in a self-sustaining loop. Without stricter checks and power to act, the office said, dangerous incidents could become more common and people could lose sight of, and control over, how the technology develops.

“Recent events have made it clear that more oversight and transparency are urgently needed,” Government Operations Agency Secretary Nick Maduros said.

Newsom said the federal government has not created meaningful AI oversight and has left regulation to companies. “California isn’t waiting until it is too late,” he said.

The California Government Operations Agency developed the report with an expert group and the California Office of Emergency Services. The advisers named in the release are Jason Goldman, a board member of the Center for Shared AI Prosperity and the first White House Chief Digital Officer; Gillian Hadfield, a professor at Johns Hopkins University; 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 United States AI Safety Institute.

The Governor’s Office said an executive order by Newsom brought the experts together to recommend updates to state AI safety laws. It also directed agencies to speed work on two laws he signed: Senate Bill 813 (McNerney), a framework for independent organizations that can assess AI systems for safety and risk, and Assembly Bill 1405 (Bauer-Kahan), which creates a state list of AI auditors and sets rules so those auditors stay independent, open about their work, and honest.

The release said a 2025 law Newsom signed already requires frontier AI developers to make their safety plans public, tell the state about certain serious safety problems, and shield people who raise alarms about grave risks. The office said Senate Bill 813 made California the first state to set a framework for certifying independent groups that can assess AI systems for safety and risk.

Newsom is asking Congress and President Donald Trump to review the state’s framework and adopt it, or to use it as a floor rather than a ceiling, the Governor’s Office said. The office did not say when any of the new recommendations would become law.