California's AI Procurement Gambit

Executive summary

California Governor Gavin Newsom's move to regulate AI through the state's purchasing power — not the legislature — is drawing sharp public skepticism, even as legal experts say the tactic may be shrewder and more durable than traditional regulation. A new pulse survey of 61 respondents finds nearly half (45.9%) oppose using government procurement to shape private markets, making it the single largest response bloc. But that reaction may be aimed at a mechanism respondents didn't fully recognize: vendors aren't legally compelled to comply — they simply lose access to the world's fourth-largest economy if they don't.

Meanwhile, a plurality (36.1%) say private companies should lead AI regulation — the top choice among four options — despite mounting evidence that voluntary industry frameworks lack enforceable teeth. A combined 52.5% favor some form of government oversight (federal or state), setting up a real tension between stated preferences and the governance options actually on the table. With the Trump administration actively working to preempt state AI laws, California's procurement gambit may be both the most legally defensible state move available and the one the public least understands.

Context

On March 30, 2026, Newsom signed Executive Order N-5-26, directing California state agencies to establish AI procurement standards within 120 days — with a hard deadline of July 28, 2026. The order targets three risk areas: illegal content generation, harmful bias, and civil rights violations. Vendors seeking state contracts must certify compliance. They are not legally required to do anything. But declining means losing access to California's government market — the contracting apparatus of the world's fourth-largest economy and home to 32 of the globe's top 50 AI companies.

The mechanism is deliberately unconventional. Bloomberg Law called it faster than legislation and harder to unwind than policy because vendors can't lobby a certification requirement before it takes effect. The order builds on California's 2023 AI executive order and slots into a broader "compliance stack" the state has been assembling — layering model safety (SB 53), transparency requirements (AI Transparency Act), and deception rules (AI Deception Act) beneath a procurement enforcement layer.

This pulse survey of 61 respondents was fielded in April 2026 — weeks after the order's signing — to capture real-time public reaction. The sample reflects general adult opinion, not a policy-specialist audience, which makes the gap between respondent perceptions and the order's actual legal mechanics a central story in the data. The survey asked four questions: two multiple-choice items on the California order and AI governance leadership, and two free-response questions on AI concerns and trust in state governments. Free-response data was scored along key attitudinal dimensions, including regulation approach and AI autonomy, using a validated LLM-assisted framework.

The federal backdrop is essential. The Trump administration's Executive Order 14365 explicitly frames state AI regulation as an obstacle to national competitiveness, and a March 2026 White House framework urged Congress to preempt state AI laws deemed burdensome. That preemption pressure is the legal environment Newsom's procurement order was designed to navigate — and understanding it changes how public skepticism about the order should be read.

Findings

Nearly half oppose procurement-as-regulation — but may be reading the wrong mechanism

The largest single bloc of respondents — 45.9% — chose "Government shouldn't use purchasing to control private markets." Only 32.8% called the order a smart workaround to federal inaction, and just 9.8% said it was good but wouldn't matter much. On its face, that looks like a public rebuke of Newsom's strategy.

But legal analysts at Bloomberg Law and Ropes & Gray describe EO N-5-26 not as market control but as a vendor certification condition attached to contract access. The state isn't telling AI companies what to build. It's deciding what it will buy. That's a distinction with enormous legal consequences — and one most respondents likely didn't have in mind when they answered. The reaction of the plurality may reflect a general wariness about government overreach rather than a considered objection to the procurement model itself.

Personality data adds a layer: respondents higher in OCEAN Agreeableness were more likely to view the order favorably (r=0.268), while those higher in Extraversion (r=−0.262) and Sociability (r=−0.251) leaned against it. The pattern suggests that comfort with cooperative, consensus-oriented approaches — rather than policy knowledge — may drive support more than ideology alone.

The private-sector self-regulation paradox

When asked who should lead AI regulation, 36.1% of respondents chose private companies — the plurality view. Federal government came second at 32.8%, followed by state governments at 19.7%. Only 11.5% selected "Other."

The preference for industry self-regulation collides hard with real-world evidence. OpenAI's proposed voluntary AI safety pact was dismissed by Anthropic CEO Dario Amodei as "PR theater" and criticized for lacking any enforceable commitments or external oversight mechanisms. A national Fathom survey found 83% of Americans want frontier AI slowed until risks are better understood — and that the public "decisively rejects a let the market sort it out approach." The gap between what this sample says it prefers and what national polling says people actually want from AI governance is one of the sharpest signals in the data.

Also notable: respondents scoring high in OCEAN Conscientiousness were less likely to select private companies as the preferred regulator (r=−0.288) — and more broadly, were less likely to pick any single option on the governance question. Those most attentive to detail and risk may be reading the governance landscape as genuinely unsettled rather than having a clear answer. The 11.5% "Other" share reinforces that signal: a non-trivial portion of the sample finds none of the options satisfying.

California is acting as a counterweight — but the feds are pushing back

Only 19.7% of respondents favor state governments leading on AI regulation, yet California is precisely doing that — and doing so against active federal resistance. Trump's EO 14365 identifies excessive state regulation as an obstacle to U.S. AI dominance and directed the Attorney General to form a task force to challenge state AI laws inconsistent with federal policy. A White House framework released March 20, 2026 urged Congress to impose federal preemption, arguing that conflicting state rules create unworkable compliance burdens.

California's procurement approach appears designed to thread this needle. Because the order conditions purchasing decisions rather than imposing binding legal requirements on vendors, it may fall outside the direct reach of federal preemption — at least as currently framed. Ropes & Gray flagged that the order even includes a provision allowing California to override federal supply chain risk designations in state contracting contexts, a direct response to the Pentagon-Anthropic dispute over national security vendor restrictions.

The broader stakes are significant. California holds roughly a quarter of global AI patents, and legal analysts suggest EO N-5-26 may join Colorado's AI Act as a national benchmark for procurement best practices. The three certification risk areas the order targets — illegal content, harmful bias, and civil rights violations — match almost precisely the top public priorities identified in national polling: child safety, accountability for harms, and verifiable standards.

Public concerns lean hard toward 'more regulation, not less'

Free-response answers to the question about general AI concerns showed a clear directional lean: the average respondent position on the Regulation Approach dimension was −0.54 on a scale from −1 (AI should be heavily regulated) to +1 (largely unregulated). That's a statistically significant lean toward the regulated pole (Wilcoxon p < 0.001), with very little expressed support for a hands-off approach.

Themes in the open-ended responses included concerns about AI-generated inappropriate or illegal content, loss of human control, and economic displacement. External research from Anthropic's survey of 81,000 workers found that roughly one-fifth voiced concern about job displacement — and that anxiety rose, not fell, among those experiencing the largest AI-driven productivity gains. Familiarity with AI appears to amplify concern rather than resolve it.

Trust in state governments to make good decisions about technology regulation was divided in open-ended responses, consistent with the split between the study's two emergent personas: "Federal-Oversight Advocates" who see state action as a necessary backstop, and "Private-Sector Optimists" who are skeptical of any single regulatory authority. Neither group is fringe — together they define the current fault line in AI governance opinion.

Conclusion

California's AI procurement order is a stress test for a question the public hasn't fully processed yet: can state-level purchasing power substitute for federal regulation when Washington is actively trying to clear the field? The survey data suggests most respondents haven't encountered that framing — and that their skepticism of the Newsom order may dissolve, shift, or harden once they understand the mechanism more precisely.

The July 28, 2026 certification deadline is the next concrete moment to watch. Whether major AI vendors comply, negotiate carve-outs, or publicly push back will reveal how much leverage California's market position actually carries. If compliance is broad, the order becomes a de facto national standard — and the template for every other large-state government that wants to shape AI markets without waiting for Congress.

For anyone tracking AI governance, the deeper signal here isn't Newsom's order in isolation. It's the widening gap between the regulatory approach the public says it prefers (private self-regulation) and the one that external evidence suggests will actually work. That gap is where the next wave of public persuasion — and political conflict — will play out.