Something shifted in the room when Eric Schmidt stepped to the podium at the University of Arizona this past May. The former Google chief executive had barely opened his remarks — a rehearsed case for artificial intelligence as the defining opportunity of a generation — before the booing started. It was not a fringe protest. It was the sound of a graduating class that had been promised a future and was being handed a liability waiver instead.

The numbers behind that moment are not soft. According to survey data cited by Semafor, more than half of Americans now hold a negative view of AI — a majority opinion that, just two years ago, would have been unthinkable to the industry's boosters. The velocity of the reversal matters. These are not people who read position papers on algorithmic bias. These are ordinary workers, students, and households who have watched AI get deployed around them, without their input, and who are drawing their own conclusions.

The standard industry response to this kind of public skepticism follows a predictable script: fear is the product of misunderstanding, the benefits will be democratized, trust will follow adoption. A fourth-edition study on AI in the workplace — the kind of recurring enterprise survey that consultancies and HR software companies commission to keep clients calm — argues along exactly these lines, pushing back on what it frames as unfounded anxieties about access and displacement. Some of those reassurances are grounded in real data. Many are not.

What the reassurance industry consistently undersells is a structural problem hiding in plain sight: the gap between who captures AI's upside and who absorbs its downside is not closing — it is widening. The workers most exposed to AI-driven job restructuring are concentrated in mid-skill, mid-wage roles. The executives making deployment decisions are not among them. This is not a technology story. It is a power story, and it has a familiar shape.

India is the clearest global pressure point right now. The country has been positioned for years as AI's great talent reservoir — a vast pipeline of English-fluent technical graduates ready to absorb the world's data-labeling, coding, and customer-service infrastructure. What that framing glosses over is that AI is now automating precisely the entry-level, export-oriented knowledge work that built India's tech middle class. The real threat isn't a talent gap. It's that careers may never get started. The first rung of the ladder is being sawed off as the climbers approach it.

Governance, the word every institution reaches for when the public gets angry, is arriving late and moving slowly. The United Nations Secretary-General has called for binding international action on autonomous weapons systems and AI risk — language that, a decade ago, would have been dismissed as science fiction. It is now standard agenda items at multilateral forums. But the gap between the urgency of those statements and the pace of actual regulation is vast. A reasonable reading of the current legislative landscape in the United States, the European Union, and most of Asia is that governments are regulating the AI of three years ago while the industry deploys the AI of next year.

The governance playbook itself may be the problem. Traditional compliance frameworks — audits, disclosure requirements, impact assessments — were designed for systems that are static and legible. AI agents, particularly those built on large language models that update continuously and operate across jurisdictions, break those assumptions at the foundation. What an audit certifies on Monday may be behaviorally different by Friday. Regulators know this. They do not yet have an answer.

What the booing at Arizona's commencement actually expressed — underneath the specific grievance about Schmidt, underneath the noise — was something more durable: a demand for accountability from people who feel they have none. The technology is not going away. The inequality of voice in how it gets deployed is the problem that leadership keeps declining to name directly. Until the workers, students, and communities absorbing AI's disruptions have genuine structural input into its governance — not advisory panels, not PR listening sessions, but real decision-making power — the gap between the public and the technology will keep widening. You cannot close a trust deficit with a better pitch deck.