Newsom Backtracks on AI Rules

California officials are known for their grandiose sense of importance. For instance, they pitch their climate change and technology regulations with an understanding that our market is so large that anything they do will have a nationwide impact. It even has a name: the California Effect. Because of our population of nearly 40 million, lawmakers believe the rules they pass will become the national standard.

And so this game is heating up when it comes to the burgeoning artificial intelligence industry, even as the San Francisco Bay Area economy and California’s capital gains-driven budget become increasingly dependent on this sector’s financial success. As Axios reported in April, the state’s latest rules follow “a familiar pattern: California acts first, companies adapt to keep doing business there, and Congress dithers, eventually ceding its role to states due to gridlock.”

The publication notes that the White House is taking a less-burdensome approach toward AI innovation, as it seeks to impose a national standard so that AI developers aren’t subject to hundreds of ever-changing rules emanating from state legislatures. Per the Trump administration’s National Policy Framework unveiled last month, “Congress should preempt state AI laws that impose undue burdens to ensure a minimally burdensome national standard consistent with these recommendations, not 50 discordant ones.”

Those recommendations are mostly on point, focusing on removing obstacles to development, promoting free speech rights, letting the courts sort through copyright issues, creating regulatory “sandboxes“ to limit government overreach, promoting workforce development, and basically erring on the side of innovation rather than government control. But the limits on states is the most significant part of that framework given the potential for mischief from Sacramento and elsewhere.

Companies of all types need predictability. As an example, the Trump administration’s tariffs have wreaked havoc in the manufacturing industry — not just because of the costs imposed by the policy itself, but because the president has continually imposed, retracted, and adjusted them. One can’t, say, invest in a new facility or set up supply chains if the government constantly changes the parameters of the taxes and regulations it’s going to impose.

Likewise with AI. Tech firms can’t set up their algorithms and models and then have to adjust them every time a state legislature imposes voluminous new rules. How can companies comply when different states might approve conflicting measures? This is, after all, a technology that isn’t contained within state (or even national) boundaries. One legislative tracker found more than 1,500 AI-related bills that have been introduced this year alone.

But, of course, California’s legislating matters most from a national perspective, with more than 20 AI bills introduced this session. As I’ve reported for The American Spectator, Gov. Gavin Newsom had previously vetoed the most far-reaching and meddlesome measures, including 2024’s Senate Bill 1047. It would have required “developers of powerful artificial intelligence models and those providing the computing power to train such models to put appropriate safeguards and policies into place to prevent critical harms,” per its legislative analysis. Those concepts are extremely ambiguous — and provided an open invitation to endless litigation.

But as my R Street Institute colleague and tech expert Mark Dalton recently explained, Newsom last month signed an executive order that includes many of the restrictions that were in the bill that he had previously vetoed. Per Dalton, the executive order creates “vast bureaucratic discretion for unelected state agency officials to decide which answers to deeply contested social and political questions are appropriate and which merit losing access to the world’s fourth-largest economy.”

The order relates only to how the state government would use and contract for AI, but it still has a fairly broad reach. The executive order calls on state officials to provide recommendations for “new certifications that may be incorporated into state contracting processes” that require companies “to attest and explain their policies and safeguards to protect public safety while preventing the misuse of their technologies.” These are largely subjective standards.

What explains the change in direction? Dalton believes it’s “a response to the Trump administration’s federal AI efforts,” but argues that “if the certifications outlined in the [executive order] were defensible in any way, they would be able to withstand legislative deliberations.” He notes, correctly, “that state procurement policy should not be driven by who occupies the White House.” Yet all Americans must continually endure the childish tit for tat from Washington, D.C., and Sacramento.

Serious policy makers would understand that the nation’s AI policy will determine our ability to compete with aggressive development efforts from other countries including China. It shouldn’t be determined by the ever-changing dynamics in various state governments. Obviously, politics always determines who makes regulatory decisions, but the rules should be consistent nationwide given that this is a national industry. That California’s governor can’t even remain consistent in his AI position year to year only reinforces that position.