California Governor-elect Xavier Becerra said he intends to work with the Legislature to turn Governor Gavin Newsom’s artificial intelligence safety recommendations into law. His statement was one of several reactions the governor’s office gathered in a release on Nov. 19. The release included only statements of support.
Newsom on Nov. 18 released recommendations developed under an executive order he signed in September, the governor’s office said. The office said the proposals would require independent oversight, set enforceable safety standards, allow dangerous AI models to be shut down in an emergency and restrict the forms of AI development that raise the greatest concern. It said President Donald Trump leaves AI companies to police themselves, while California is leading.
Becerra said California already has the nation’s first frontier AI safety law, though not yet the strongest one. Frontier AI refers to the most advanced AI systems. He said he wants to move quickly in January to put standards, independent verification, full incident reporting and shutdown authority into law. “The companies building this technology do not get to grade their own safety,” Becerra said.
Thomas Woodside, co-founder of Secure AI Project, listed what he saw in the report. These included mandatory embedded evaluators, state power to order corrective action in emergencies, mandatory minimum safety standards, and limits on the computing power used for automated AI research. He said Newsom and the Legislature need to act now.
AVERI, a nonprofit that works on third-party auditing of frontier AI, recommended that Newsom call a special session or use existing powers to act. The release did not say whether he plans to do either. AVERI also said a company that invests heavily in AI research and development should count as a large frontier developer regardless of its revenue.
Charlie Bullock, senior policy director at LawAI, said the report leaves some design questions to the Legislature, including who pays for and selects the independent verifiers. Andrew Freedman wrote that turning the report into law and rules will raise debates over state and federal power, and over logistics, timing and costs. Jay Shooster, executive director of Citizen Safe AI, and Jason Wolfe also voiced support.
The release did not give a timeline for legislation or say what the proposals would cost.
