With governance at a crossroads, CA tries to seize control of its AI destiny

California has always been a trend setter, and not just in movies and music. The state has often taken a bold role in establishing policies and regulations that push the boundaries established by other jurisdictions – including the federal government. 

The governance of artificial intelligence is shaping up to be no exception.  

This week, California Governor Gavin Newsom issued an executive order that aims to preserve his state’s right to set its own AI policy, squaring up with the Trump Administration in light of recent efforts to centralize and deregulate oversight of AI technology instead. 

A call for more rigorous standards for AI developers

The executive order directs the California government to strengthen control of how AI technologies are procured and used within the state by taking several actions, including creating a new vendor certification framework that requires vendors to prove they have certain safeguards in place around algorithmic bias, prevention of harmful or illegal content, and the protection of civil rights and liberties. 

The order also directs state agencies to establish new procurement rules and contracting requirements, define what “responsible AI” really means in practice, and create specific evaluation criteria for vendors that hope to work with state entities. 

By requiring vendors to adhere to certain safeguards in order to do business with the world’s 4th largest economy, Newsom hopes the effects will spill over into the private sector in California and elsewhere, thereby raising the floor for AI tools as a whole. 

“California’s always been the birthplace of innovation. But we also understand the flip side: in the wrong hands, innovation can be misused in ways that put people at risk,” Newsom said in a press release. “California leads in AI, and we’re going to use every tool we have to ensure companies protect people’s rights, not exploit them or put them in harm’s way.” 

Asserting independence from federal regulatory decision making

The order comes just weeks after the Trump Administration issued its National Policy Framework for Artificial Intelligence that calls upon Congress to pass a law that overrides state laws around AI and creates a single national plan instead.  

The framework is itself a follow-on to a December 2025 federal executive order that tries to limit state-level lawmaking around AI by initiating lawsuits and restricting funding opportunities for states that “require AI models to alter their truthful outputs, or…compel AI developers or deployers to disclose or report information in a manner that would violate the First Amendment or any other provision of the Constitution.” 

Despite the White House’s insistence on this point, Congress has already overwhelmingly voted against a freeze on state-level AI legislation. 

Nevertheless, California is doubling down its point about state’s rights by setting up a vendor evaluation framework that does not reference or necessarily even acknowledge federal standards, and, in the press release, claiming the authority to “separate its procurement authorization process from the federal government’s if needed” to protect the state’s perceived interests. 

Get ready for protracted wrestling for the future of AI governance

Both the federal executive order and California’s counterplan set up a battle that is likely to play out in various courts, with the Republican administration, somewhat ironically, trying to push a unified, federalist perspective while deep blue California is on the side of state-level self-determination. 

The prize is the essence of AI governance itself.  

While this Administration isn’t going to last forever, and the next resident of the White House may have a different view entirely on how to govern AI development, there’s no getting around the fact that the next two to three years of development in artificial intelligence are going to be intensely formative.   

What happens during this time period will be crucial for shaping the next decade or two of the AI-powered ecosystem, and getting governance right will be non-negotiable. 

There are plenty of risks to consider. For one thing, Trump’s governance approach isn’t just about taking a strong, coordinated, top-down approach to set standardized guardrails for everyone to prevent fragmentation. Instead, it’s designed to limit states from doing much in the way of lawmaking while actively deregulating what guardrails do (or did) exist so that developers have a free hand to “innovate” without interference at any other regulatory level. 

However, individual states taking too much of AI governance into their own hands could produce the opposite problem: a patchwork of impossible-to-navigate regulations that make it difficult for vendors to operate across state lines while meeting every single applicable policy. 

The solution, of course, would be a common-sense blend of firm federal guardrails that still retain flexibility for states to implement rules that align with their unique priorities and protect their specific populations. But the odds of achieving such a balance in the current economic and political climate seem fairly remote. 

Instead, AI vendors and adopters will need to keep a careful eye out for lawsuits, legislation, more EOs, and other policymaking at both the state and federal levels that could change how they operate.  

It won’t be easy to keep track of how this will all unfold, and the outcome is still uncertain as the courts haven’t yet really gotten to grips with specific cases yet. The best thing to do right now is for vendors and implementors to hold themselves to the highest shared standards of governance to best protect the users of AI technologies. 

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at [email protected].