Becerra and Hilton Withhold Full Backing for Newsom’s AI Safeguards

California’s next governor may inherit a sharper fight over AI safety than either leading candidate is ready to settle. Xavier Becerra and Steve Hilton both say AI needs regulation, yet neither has fully endorsed Gov. Gavin Newsom’s proposed path toward independent evaluations and emergency model shutoffs.

Newsom’s September 18 executive order starts a policy process rather than imposing new duties on AI developers. It speeds implementation of California’s recently enacted laws on independent AI verification and auditor standards, while asking national experts for recommendations on stronger rules.

The options include placing independent verification organizations in frontier AI companies’ labs for regular audits and evaluations. Experts are also to consider how California could advance an emergency “kill switch” for frontier models and have an outside organization verify that it works over time.

Hilton has called AI-risk regulation a “very urgent priority” and said he is interested in external safety monitoring. But he said policymakers should not turn recent alarm over AI risks into what he described as “really bad regulation.” He supports convening a panel to recommend safety policy, even as he rejects Newsom’s most concrete proposed emergency control.

Becerra’s campaign took a less specific position. It said California cannot accept technology that poses catastrophic risks to public safety and said Becerra would set firm guardrails around dangerous AI capabilities and systems operating beyond human oversight. Asked whether he would mandate kill switches and outside monitoring, the campaign did not say yes.

The experts’ recommendations are due November 16, a schedule that could give Newsom time to pursue a special legislative session before leaving office. California already requires large AI-model developers to disclose how they manage catastrophic risks and report qualifying safety incidents. The new review asks whether the state should add outside scrutiny and an emergency response mechanism to that disclosure-and-reporting system.