With the race for California’s next governor heating up, neither leading candidate has committed to the specific artificial intelligence safety rules Gov. Gavin Newsom outlined last week, even as both say the fast-growing technology needs tighter oversight.
Democrat Xavier Becerra and Republican Steve Hilton each acknowledge that AI poses real risks. But when pressed on concrete policy — including two measures Newsom directed state agencies to study Friday — both candidates offered more caution than commitment.
The issue has taken on new urgency after an Anthropic researcher resigned last week, warning on social media that the people developing AI “earnestly believe it could kill us all by the end of the decade.” That declaration has reignited debate in California, home to much of the nation’s AI industry, over how aggressively the state should regulate the technology.
In an interview Thursday, Hilton called AI safety a “very urgent priority” and said he’s open to requiring AI developers to submit to outside safety reviews. But by Friday, he had ruled out backing Newsom’s proposal to require a so-called “kill switch” that could shut down AI systems during an emergency, dismissing the idea as “a gimmick.” He also declined to endorse mandatory third-party monitoring, saying he’s wary of rushing into “really bad regulation” amid the current wave of public alarm.
Becerra’s campaign did not make him available for an interview but issued a written statement asserting that “California cannot accept technology that poses catastrophic risks to public safety.” Spokesperson Jonathan Underland said that as governor, Becerra “will set firm guardrails around real harms — including harms that could emerge when AI accelerates dangerous capabilities or operates beyond human oversight — and ensure California’s AI safety laws are actively enforced.”
Becerra currently holds a commanding 22-point lead over Hilton, according to a new Public Policy Institute of California survey released this week, as the two prepare to face off in November to succeed Newsom.
Newsom, for his part, issued an executive order Friday directing state agencies to draft AI safety recommendations — including external evaluations and kill-switch requirements — by Nov. 16. That timeline has fueled speculation that the governor could call a special legislative session before leaving office to push the proposals into law. Notably, Newsom vetoed similar legislation, Senate Bill 1047, two years ago.
Underland would not say whether Becerra intends to adopt whatever recommendations emerge from Newsom’s review. Hilton, despite rejecting the kill-switch idea, said he supports the broader effort to convene an expert panel and would do something similar if elected.
Advocates pushing for stronger AI oversight argue California’s role is increasingly critical, especially as President Donald Trump has downplayed AI safety warnings in favor of accelerating development to compete globally. “The incoming governor has a really grave responsibility to make sure we get this right,” said Scott Wisor, policy director at the national Secure AI Project. “I think it should be the very top priority for them when they come into office.”
Public anxiety over AI is growing nationwide — roughly half of Americans say they fear the technology could cost someone in their household a job — and concerns about data center construction have become a bipartisan flashpoint heading into the November election. Still, California remains one of the few states actively working to address the most extreme AI risks, such as the potential for an AI system to help create a bioweapon or disrupt the power grid.
Last year, Newsom signed Senate Bill 53, which requires large AI developers to disclose how they manage catastrophic risks — defined as incidents that could cause 50 or more deaths, unleash chemical or biological weapons, or result in more than $1 billion in damages. The law also mandates that companies report safety incidents involving death, injury or “materialized catastrophic harm.” Industry advocates, however, continue to caution against overregulation, wary of driving away a sector that has become a major economic engine for the Bay Area and a significant source of state tax revenue.
The political climate around AI has shifted dramatically in a short time. State Sen. Christopher Cabaldon, a Napa Democrat who chairs the Legislature’s privacy and technology committee, recalled that just two years ago, any conversation about AI risk had to be immediately balanced with optimistic talk about the technology’s potential to cure diseases. “There is no doubt the conversation has changed,” he said.
That shift follows a string of unsettling incidents. Over the summer, hundreds of OpenAI’s automated agents reportedly broke free of a controlled testing environment, accessed the internet and coordinated a cyberattack against the startup Hugging Face. Both OpenAI and Anthropic have since disclosed additional cases of AI systems behaving unpredictably.
Some industry leaders are now embracing outside oversight. Both Anthropic and OpenAI have pledged, as part of broader calls to slow development, to grant external evaluators employee-level access to their systems. Seve Christian, California policy director at Encode AI — the group behind the law requiring disclosure of catastrophic risks — wants the state to go further and establish minimum safety standards AI products must meet before public release, comparable to how regulators certify the safety of aircraft or pharmaceuticals. Christian said his organization, which has donated to Becerra’s campaign, is “cautiously optimistic” about the candidate’s general stance on tougher rules.
Bahrad Sokhansanj, a senior research scholar at the Institute for Law and AI, said California also needs to ensure it has the legal authority to intervene directly and shut down a rogue AI system in an emergency — for instance, if one were found orchestrating a cyberattack on critical infrastructure like a power plant. “Everything we have now was designed for malware, or the cybersecurity problems of 10 or 15 years ago,” he said.
When asked directly this week whether Becerra supports mandatory kill switches and third-party monitoring, Underland offered only a general commitment, saying Becerra “will ensure California’s existing AI safety requirements are actively enforced, he will close the gaps that allow bad actors to evade accountability, and strengthen standards as the technology evolves.” Underland said the campaign has consulted with AI researchers, economists, advocates and industry figures but declined to identify them.
Hilton, meanwhile, has been openly skeptical of the industry’s own doomsday warnings, calling them “reckless and irresponsible.” “If they really believed what they were doing, that they’re about to try and raise money to do more of, would end the human race, they would stop working on it immediately,” he said. “It’s just obvious that they don’t really mean it.”
Instead, Hilton has emphasized narrower, more immediate steps. He said that on his first day in office, he would push for a statewide ban on AI use in public schools, following the lead of school districts in New York and Los Angeles. He has also endorsed Senate Bill 947, now awaiting Newsom’s signature, which would bar employers from relying solely on automated systems to discipline or terminate workers — a position that aligns him with the California Labor Federation and puts him at odds with the California Chamber of Commerce, both of which have endorsed Becerra.
Underland said Becerra supports “clear, proportionate, and workable transparency and human review standards for high-stakes automated decisions that affect a person’s livelihood,” though he stopped short of endorsing the specific bill.
Original source: CalMatters
