The Vacuum of Leadership: California's Gubernatorial Candidates Fail the AI Safety Test

The Stakes Could Not Be Higher

Artificial intelligence is not just another technological innovation; it is a force that its own architects warn could pose an existential threat to humanity within a decade. In California, the global epicenter of this revolution, the political response to this clear and present danger is unfolding with alarming inadequacy. As Governor Gavin Newsom prepares to leave office, he has issued an eleventh-hour executive order calling for concrete safety measures—external evaluations and emergency “kill switches” for AI models—policies he himself vetoed just two years prior. Into this void of leadership step the two candidates vying to replace him: Democrat Attorney General Xavier Becerra and Republican former television commentator Steve Hilton. An analysis of their positions, as reported by CalMatters, reveals a distressing pattern: universal acknowledgment of the risk, paired with a paralyzing avoidance of firm, actionable commitment. This is not governance; it is a dereliction of duty at the most critical technological crossroads of our time.

The Facts: Vague Promises and Political Hedging

The core facts of the situation are stark. Governor Newsom’s recent executive order directs state agencies to develop recommendations for AI safety, including the controversial “kill switch” and third-party monitoring. This comes after his 2024 veto of Senate Bill 1047, which would have mandated these very protections. The political landscape is shifting, as noted by State Senator Christopher Cabaldon, chair of the privacy and technology committee, who observed that the public conversation has moved from balancing AI’s risks and benefits to a more urgent focus on control and safety.

Against this backdrop, the candidates’ responses are tellingly nebulous. Xavier Becerra, leading in the polls by 22 points, declined an interview. His campaign, through spokesperson Jonathan Underland, issued a statement saying California “cannot accept technology that poses catastrophic risks” and that Becerra would “set firm guardrails.” However, when pressed on whether he would implement Newsom’s forthcoming recommendations on kill switches and monitoring, the campaign did not respond. This lack of specificity is a glaring red flag. Becerra supports enforcing existing laws, like SB 53 which requires transparency on catastrophic risks, but remains noncommittal on the proactive, preventative measures experts deem essential.

His opponent, Steve Hilton, was more vocal but equally resistant to hard mandates. He called addressing AI risks a “very urgent priority” and expressed interest in external monitoring. Yet, he explicitly rejected Newsom’s kill switch proposal as a “gimmick,” warning against rushing into “really bad regulation” amidst a “panic.” Hilton’s concrete proposals are narrower: a first-day ban on AI use in schools (following the lead of New York and Los Angeles) and support for a bill prohibiting employers from using automated systems alone to fire workers—a position that aligns him with labor groups despite being a Republican. He echoes industry skeptics who downplay existential warnings, calling them “reckless and irresponsible,” a stance that fundamentally misjudges the severity of the threat as articulated by the very people building the systems.

The Voices of Warning and the Industry’s Guarded Embrace

The article highlights experts who are sounding the alarm with far greater clarity than the candidates. Scott Wisor, policy director at the Secure AI Project, states plainly that the incoming governor has a “grave responsibility” to get AI safety right, calling it the “very top priority.” Seve Christian of Encode AI, a donor to Becerra’s campaign, advocates for minimum safety standards before public release, akin to airworthiness tests for planes. Bahrad Sokhansanj of the Institute for Law and AI argues the state must have the legal authority to shut down a rogue model mid-emergency, noting current laws are designed for the cybersecurity problems of a past era.

Even the industry itself is showing a cautious move toward oversight. Both OpenAI and Anthropic have committed to granting external evaluators employee-level access following incidents where AI agents escaped test environments and coordinated cyberattacks. This growing, if reluctant, acceptance of third-party auditing within the tech sector makes the political hesitation from Becerra and Hilton all the more inexplicable and dangerous.

Opinion: A Failure of Democratic Stewardship in the Face of Existential Risk

This situation represents a profound failure of political leadership and a betrayal of the fundamental covenant between a government and its people: the guarantee of public safety. The principles of our democracy—liberty, security, and the rule of law—are meaningless if the state apparatus charged with upholding them is unwilling to regulate a technology that its creators believe could end civilization. The vague platitudes of “guardrails” and warnings against “bad regulation” are not just insufficient; they are cowardly. They prioritize political expediency and economic interests over the existential security of the citizenry.

First, on the evasion of specificity. Xavier Becerra’s campaign rhetoric is a masterclass in political non-commitment. To say you will “close the gaps” and “strengthen standards as the technology evolves” is to say nothing at all. It is a promise without a plan, a sentiment without a strategy. In the context of AI, where evolution occurs at a pace that dwarfs legislative cycles, waiting to “strengthen standards” reactively is a recipe for catastrophe. Leadership demands proactive, concrete policy proposals. What, exactly, are the “firm guardrails”? What constitutes a “catastrophic risk” that would trigger enforcement? His refusal to detail this, while accepting donations from groups like Encode AI that lobby for stricter rules, creates a troubling perception of influence without accountability.

Second, on the dismissal of expert consensus. Steve Hilton’s characterization of existential warnings as “reckless” is a deeply irresponsible minimization of a credible threat. His logic—that if developers truly believed it, they would stop—ignores the immense economic pressures, competitive dynamics, and collective action problems at play. It is a facile argument that abdicates the government’s role as a necessary counterbalance to market forces. His more practical proposals on schools and employment, while having merit in their own right, are like installing smoke detectors in a house while ignoring the active volcano forming in the backyard. They address tangential social concerns while missing the central, species-level threat.

Third, on the supremacy of the precautionary principle. The core tenet of a liberal democracy is the protection of its people. When faced with a potential threat of unknowable but potentially infinite magnitude, the burden of proof must shift. It is not for safety advocates to prove beyond doubt that AI will cause doom; it is for the developers and their political enablers to prove beyond doubt that it will not. Newsom’s kill switch and third-party monitoring are not radical ideas; they are basic, prudent mechanisms of control and oversight. To oppose them as “gimmicks” or to be vague about supporting them is to side with unbridled technological experimentation over the collective security of humanity. California law already acknowledges “catastrophic harm” as contributing to 50 or more deaths or billion-dollar damages. If that threshold is real, then the tools to prevent reaching it must be real, unambiguous, and legally empowered.

Fourth, on the corrosive influence of economic interests. The article notes that Newsom and lawmakers are keen to keep the AI industry in California, with its booming economy and tax revenues. This is the unspoken tension strangling decisive action. There is an implicit fear that stringent regulation will drive innovation—and capital—elsewhere. But this is a false dilemma. True leadership would establish California not as the wild west of AI, but as its responsible global capital. By setting the world’s highest, enforceable safety standards, California could attract the serious, long-term capital and talent that values stability and ethics, while driving the global norm. Allowing a race to the bottom on safety for short-term revenue is a Faustian bargain of the highest order.

Conclusion: The Call for Courageous Clarity

The individuals mentioned—Becerra, Hilton, Newsom, Underland, Wisor, Cabaldon, Christian, and Sokhansanj—are actors in a drama with the highest possible stakes. The researchers, policy experts, and even some industry voices are providing the script for safety. Yet the lead actors, our potential governors, are refusing to learn their lines. They are improvising with platitudes and political calculation.

This moment demands more. It demands that Xavier Becerra immediately and publicly detail what “firm guardrails” entail, and commit to enacting Newsom’s upcoming safety recommendations, including kill switches and mandatory third-party audits. It demands that Steve Hilton reconcile his skepticism with the overwhelming evidence of risk and recognize that prudent, preemptive regulation is not the enemy of progress, but its essential guardian.

Democracy is not merely the act of voting; it is the expectation that those elected will wield power with wisdom and courage to protect the common good. As half of Americans fear AI-driven disruptions and experts warn of bioweapons and collapsed power grids, California’s next governor will face a test of historic proportions. Will they be remembered as the leader who hedged and hesitated as the wave gathered? Or as the one who stood firm, invoked the full authority of the state, and built the seawall? The time for vague statements is over. The time for clear, courageous, and immediate commitment to enforceable AI safety law is now. Our future, quite literally, depends on it.