The AI industrial complex is drafting a generation of graduates. It is true that the current generation of large language models works, in well-defined domains, at substantial cost, and it is true that the capital expenditure behind the data centers is a real-economy infrastructure event that ratepayers and consumers will ultimately balance on their ledgers. The trouble is that the commencement-stage enthusiasm for these systems, deployed by platform-companies’ policy proxies and tech-celebrities on a college-quadrilateral circuit this May, is less a recognition of engineering progress than a coordinated attempt to normalize their inescapability before the workforce has accumulated enough leverage to say no.
The friction at Glendale Community College was far more than an embarrassing technical failure. An automated system called a cloud API to read names, suffered a prompt-engineering failure, and read the wrong names aloud—while some students’ names were skipped entirely, leaving them standing on stage without even being called. And here it is worth being precise about what an “AI reader” actually is, because the public discourse has the misleading habit of treating it as a novelty rather than a continually-tuned set of weights serving a continually-revised objective function—they are just a software pipeline parsing a PDF and hallucinating on the first call—so when the API times out, the college president had to stand on stage and tell a room of students that their ceremony was being outsourced to a server rack. It was a perfect, performative microcosm of the current digital-policy moment: the institution offloaded a basic human task—one that constitutes the entire point of a graduation ceremony—onto a cheap automated tool and ruined the moment for the very students the college supposedly exists to serve.
At the University of Central Florida, the real estate executive Gloria Caulfield described the technology as the next industrial revolution and struck a chord the students had no patience to explain. At Middle Tennessee State University, the record executive Scott Borchetta told graduates the technology was rewriting production, a statement that elicited a booing sharp enough for Mr. Borchetta to counter that everyone had to simply deal with it. At the University of Arizona, the former Google CEO Eric Schmidt repeated the claim that artificial intelligence will shape the world and met the same acoustic rebuttal. What these speakers fail to grasp as they lecture from the podium is that the audience they are addressing has spent four years watching the same machine-learning apparatus prune their job prospects and automate away the very internships they were told would lead to their first careers.
The graduates at American University and the University of Denver were not rejecting the technology in the abstract. They were rejecting the attempt to position the technology as a force of nature—a “tool” that “rewrites production as we sit here”—rather than a product built by labor under management decisions made for documented reasons. They were responding correctly to a raw power play: the AI industrial complex is deliberately rolling the “AI is coming for your job” narrative directly against the most vulnerable cohort in the economy to preemptively hollow out their bargaining power, positioning themselves as the mandatory chokepoint between the graduating class and the workforce.
Students were right to connect the noise to the wealth extraction. As one student told NPR, the technology is making billionaires richer while depleting the environment, with data centers drawing water and electricity from municipalities that cannot subsidize the demand. The Quinnipiac poll, conducted by associate professor Chetan Jaiswal, corroborated the students’ structural assessment, finding that 81 percent of Gen Z believes AI will decrease job opportunities, while only 5 percent of Americans feel AI development is led by groups that represent their interests. Jaiswal captured the shift: “People are not rejecting AI, but people are asking questions now since the initial AI fever is gone.” This is not anti-technological hysteria. It is a rational assessment of a system being rolled out without regulatory guardrails, without safety standards for automated decision-making, and without any transparent accounting of the energy and water it consumes. These graduates are not looking for a lecture on how to be a “centaur” for their corporate masters; they are looking for a labor market that hasn’t been deliberately gutted.
The playbook of extracting the surplus that the existing workforce has built, locking the workforce in by raising the cost of leaving, and doing it to the next class of suppliers is not new to families of this trade, though the mechanism of the digital lock is novel in its opacity. A millwright used to measure twice and cut once, checking torque specs against a repair manual that the machine’s owner had a right to read. The current architecture choice is the same extraction pattern under a different label: a platform-companies’ management decision to embed cheap firmware in replacement software, enforce a digital lock via anticircumvention statutes, and bill the operator for the right to run the asset they already own.
There is a public consultation open at the Federal Communications Commission and the Federal Trade Commission regarding the deployment of large language models in education and employment screening. The consultations are unlikely to produce a bill that fundamentally alters the vertical integration of the cloud layer, the ad-tech stack, and the AI inference markets. None of this is a reason not to submit. Deadlines are the only part of regulatory processes that the platform companies actually respect, and submissions are the only part of regulatory records that subsequent governments have to read. Until the platforms are subjected to actual antitrust constraint, the graduates at Arizona, at Middle Tennessee, at Glendale have every reason to keep booing the people trying to sell them the wreckage of their own professions. There is a tradesman’s maxim that works as well on server racks as it does on bar mills: you can pay extra for the privilege of being denied the thing you already built.