AI Governance Trust Crisis

Executive summary

A federal trial over who controls AI just ended without answering the question that matters most — and the American public knows it. The Musk v. Altman case concluded in May 2026 on a procedural technicality, leaving OpenAI's for-profit conversion legally intact and the underlying governance debate entirely unresolved.

A survey of 137 Americans fielded during the trial found that nearly four in five respondents — 79.5% — are primarily worried about either AI companies prioritizing profits over safety (44.5%) or too much power concentrated in too few hands (35.0%). Trust in tech billionaires skews sharply negative. And when asked who should govern AI, no single authority — not scientists, not companies, not regulators — commands even three in ten votes.

The trial's outcome didn't settle anything. It may have made the anxiety worse.

Key takeaways:

• 44.5% cite profit-over-safety as their top AI concern — the single largest response
• 79.5% of concerns cluster around governance failures, not funding shortfalls
• Independent scientists lead governance preferences at 28.7%, but no option breaks 30%
• Trust in tech billionaires clusters at the lowest end of the scale
• The EU AI Act's enforcement machinery doesn't activate until August 2026

Context

For three weeks in May 2026, a San Francisco federal courtroom became the unlikely arena for the biggest governance question in tech: Can you build transformative AI under a nonprofit mission, or does the money required make that impossible?

Elon Musk sued OpenAI and CEO Sam Altman, alleging the organization betrayed its founding charitable purpose when it restructured its for-profit subsidiary — and enriched insiders in the process. Altman testified that the co-founders believed "no single person should control AGI." Musk sought disgorgement of up to $150 billion and the removal of Altman and OpenAI president Greg Brockman. OpenAI countered that Musk launched rival company xAI and was motivated by competitive interests, not principle.

The jury never ruled on those claims. A unanimous verdict found Musk had waited too long to sue — his claims were time-barred under a three-year statute of limitations. OpenAI's conversion to a public benefit corporation, completed during the trial, stood. Musk announced an appeal, calling the ruling "a calendar technicality."

The survey captured public sentiment during this window — 137 U.S. adults responding to four questions about AI governance, trust, and funding. The questions were designed to surface which concern dominates the public mind: safety versus profit, power concentration, or the sheer cost of keeping frontier AI research alive. Responses were collected with personality trait data (OCEAN model) for a subset of participants, enabling correlation analysis between individual psychological profiles and governance attitudes.

The backdrop matters: frontier AI training costs have grown at 2.4 times per year since 2016, with the largest runs projected to exceed $1 billion by 2027. Both Musk and Altman agreed in court that building advanced AI requires massive capital. The question the trial never answered — and that regulators haven't yet addressed — is who gets to decide the terms under which that capital flows.

Findings

Safety fears and power anxiety crowd out every other concern

The numbers from the first question are stark. When given four options about AI development, nearly half of respondents — 44.5% — said their primary worry is AI companies prioritizing profits over safety. Another 35.0% flagged too much power concentrated in a few hands. Together, those two responses account for 79.5% of all answers. Only 13.9% said insufficient funding was the main problem.

That 13.9% figure deserves attention precisely because it's so low. The economic case for the for-profit pivot is real: AI training costs have grown at 2.4 times per year since 2016 and are on track to exceed $1 billion per run by 2027. OpenAI's restructuring was, in significant part, a response to that trajectory. Yet fewer than 1 in 7 respondents chose funding shortfalls as their top concern. The public has absorbed the message that AI is expensive — and decided that's not the thing to worry about.

What they are worried about is whether the people spending that money can be trusted. A March 2025 Change Research survey found 62% of U.S. voters favor strict government safety rules for AI even at the cost of innovation, with majority support across party lines. The survey data here tracks that national signal closely.

No single authority earns public confidence to govern AI

The governance question produced the most fragmented result in the survey. Asked who should have the most say in AI development, respondents spread their answers almost evenly across four choices: independent scientists (28.7%), tech companies (24.3%), government regulators (23.5%), and a substantial "Other" category (23.5%).

No option broke 30%. That's not indecision — it's a signal that the public doesn't see a legitimate authority already in place. The nearly quarter of respondents who chose "Other" are effectively saying: none of the above. The Cicero Institute's 2025 national AI poll found a parallel result: scientists and researchers topped trust rankings at 38%, while government scored just 6% as the entity most trusted to develop AI. Forty-four percent supported creating an independent government council to advise on AI laws — a structure that doesn't currently exist.

One notable pattern emerged in the personality data: respondents with higher Prism Influence scores — a measure associated with social persuasion and leadership orientation — were more likely to select independent scientists as their preferred governance authority. Higher Influence scores also correlated negatively with choosing government regulators. The people most attuned to how power operates appear most skeptical that existing power structures should run AI.

Trust in tech billionaires runs near zero

Free-response data on trust in figures like Musk and Altman skewed heavily toward the bottom of the scale, with responses clustering at 1 and 2. This wasn't a nuanced distribution — it was a pile-up at the low end.

Personality data adds texture to that finding. Higher scores on the OCEAN Neuroticism dimension — a measure of emotional reactivity and tendency toward anxiety — correlated negatively with trust in tech billionaires (r = -0.274). Higher Agreeableness correlated positively (r = 0.270), suggesting that more cooperative, trusting individuals extend some benefit of the doubt to AI executives, while more emotionally reactive respondents do not.

The Neuroticism connection has external support: a study in the European Journal of Sustainable Development found Neuroticism correlates with AI anxiety at r = 0.301, with illusory beliefs about technology mediating the relationship. In other words, anxious respondents aren't just reacting to facts — they're processing AI through a framework already primed for threat detection. For communicators, that means factual reassurance alone is unlikely to move this cohort.

There's an irony embedded in the trial testimony itself: Sam Altman stated that the co-founders believed "no single person should control AGI" — yet the organization he leads is now a public benefit corporation whose governance critics describe as concentrating effective control. The gap between stated principle and perceived reality is exactly where public distrust lives.

The trial ended. The question didn't.

The Musk v. Altman verdict resolved nothing substantive. A unanimous jury found Musk's claims time-barred — meaning the court never evaluated whether OpenAI actually betrayed its nonprofit mission. OpenAI's conversion to a public benefit corporation proceeded regardless. Musk appealed.

The California Attorney General's office is separately investigating the conversion, citing OpenAI's articles of incorporation, under which assets are "irrevocably dedicated to its charitable purpose." Critics argue that allowing a nonprofit to repurpose those assets could set a precedent for other startups to exploit tax-exempt status before converting to for-profit structures.

Meanwhile, the governance gap is structural. As a Just Security analysis noted, when Anthropic completed a Claude model capable of exploiting zero-day vulnerabilities across major operating systems and browsers, no regulator had authority to demand it not be released. The company chose restraint voluntarily. The EU AI Act's enforcement tools don't activate until August 2026, and even then, mitigation standards remain vague. With 79.5% of this survey's respondents expressing governance-related concerns — and no authoritative resolution in sight — public anxiety has nowhere to go but up.

Conclusion

The Musk v. Altman trial was supposed to force a reckoning over whether profit motives can coexist with AI's original public-benefit mission. Instead, it ended on a technicality — and the reckoning got deferred to regulators, attorneys general, and an anxious public with no clear authority to turn to.

What this survey makes plain is that the public has already rendered its verdict on the underlying question, even if the court didn't. Nearly four in five respondents are primarily worried about governance failures — profit over safety, power in too few hands — not about whether AI labs have enough money to operate. The economic case for the for-profit pivot didn't land. Trust in the people running these systems is near the floor.

Watch three things in the coming months: the California AG's investigation into OpenAI's asset conversion, which could establish precedent for how AI nonprofits are allowed to restructure; the EU AI Act's enforcement activation in August 2026, which will be the first real test of whether regulatory frameworks can keep pace with capability development; and the appeal Musk filed, which keeps the substantive governance question alive in court even if the statute-of-limitations ruling stands.

The public isn't waiting for those outcomes to form opinions. The anxiety is already baked in — and it's growing.