The University’s Soul Has Been Uploaded
As universities embrace generative AI, they are allowing Big Tech to reshape teaching and learning without first answering a fundamental question: Does this technology belong in the classroom at all?
Universities are making a consequential concession: allowing privately owned generative AI systems into the very processes through which students learn to read, write, reason, and exercise judgment. Their adoption is increasingly presented as inevitable. AI exists, so higher education must adapt. Under this logic, adoption is expected, while refusal requires an explanation. But technological availability is not educational justification. Before reorganizing teaching and learning around these systems, universities should answer the question that advocates of adoption too often avoid: Why should AI be there at all?
The language of inevitability obscures who is responsible. A recent Forbes headline declared, “MIT Says AI Is Forcing a Rethink of College Itself.” AI is not forcing universities to purchase licenses, revise courses, shift budgets, or surrender influence over educational infrastructure to private vendors. People make those decisions. They should have to explain and defend them. Who was consulted? What alternatives were considered? And whose interests are being served? California State University signed a $16.9 million OpenAI agreement in 2025. A statewide student leader said her association wasn't consulted; officials cited an advisory committee with student and faculty members.
A university exists to cultivate independent thought, critical judgment, curiosity, and the ability to work through difficult questions. Reading closely, developing an argument, recognizing errors, and revising one's thinking are not incidental steps on the way to an answer. They are part of what education is for. What happens when students are encouraged to rely on systems that can perform much of this work for them?
Public universities owe Big Tech no duty of adaptation. Their obligation is to students, workers, and the public.
The available research offers no warrant for treating adoption as an educational obligation. A 2026 meta-analysis of 35 studies reported moderate gains across varied educational settings. A review of 67 higher-education studies found better outcomes when AI was used within carefully structured, inquiry-based teaching. Those findings deserve consideration, but they do not settle whether AI belongs in the learning process. In the latter review, stronger results depended on deliberate human instruction, and neither review establishes that AI reliably develops lasting, independent learning or provides benefits that effective teaching without it cannot achieve.
Better assignments are not necessarily evidence of better learning. A polished paper produced with ChatGPT does not show that its author has become a better writer. Short-term performance measures tell us little about what students may be losing when they outsource reading, writing, and reasoning. Assisted performance is not intellectual development, and making learning easier does not necessarily make education better. Nor can learning-outcome studies settle the questions that institutions routinely sidestep: Who should own educational infrastructure, what student work may be collected, whose labor may be automated, and whether students and faculty can refuse participation. Those are ethical and political decisions, not matters of test scores alone.
Yet narrowly demonstrated benefits are being used to support sweeping institutional ambitions. Universities must consider training, faculty workload, verification, privacy, dependence on proprietary systems, and what happens when a vendor changes its prices or rules. Once courses depend on a platform, leaving can become disruptive and expensive. Public money finances the infrastructure, faculty labor makes it part of everyday teaching, and the resulting dependence can give vendors greater leverage over academic decisions. Public institutions should disclose the full costs, contractual obligations, data protections, and exit plans before committing resources and educational authority. These are questions of public accountability, not merely technical procurement.
For faculty, promised efficiency can mean additional labor. Professors redesign assignments, check unreliable citations, devise oral defenses, interpret shifting policies, and struggle to determine whether students have done the intellectual work being assessed. That verification work consumes time that could otherwise go toward mentoring, substantive feedback, and sustained classroom discussion—costs rarely visible in claims of efficiency. At the same time, AI is promoted to automate feedback, lesson planning, and other forms of professional judgment. Faculty are asked to help install the very systems that may eventually diminish their autonomy.
Students face another danger in the language of “personalized learning.” Systems built around continuous assessment, behavioral prediction, and linked records can turn intellectual development into an expanding data profile. Students should be free to misunderstand, hesitate, write badly, and change their minds without every misstep becoming a permanent record. These are conditions for learning, not inefficiencies to be recorded and corrected. And when students' writing, questions, and personal circumstances pass through commercial systems, privacy cannot be an afterthought. Students deserve to know what is collected, who can access it, how long it is kept, and whether they can decline without jeopardizing their education.
Universities normalize tools that make it harder to determine whether students worked independently, then turn to monitoring technologies to establish whether they did. Instead of asking whether the original technology undermines assessment, institutions risk treating surveillance as the solution.
Beyond the classroom, students face demands to “upskill” for an AI-driven economy. Employers may promote AI proficiency while using automation to cut labor costs, reduce professional discretion, and concentrate control. The political question of who controls automation and receives its benefits is being recast as an individual problem of employability. Universities should not simply prepare students to accommodate those decisions. They should equip students to ask whose knowledge AI absorbs, who benefits, and what protections workers need—including collective bargaining and enforceable rights.
Public universities owe Big Tech no duty of adaptation. Their obligation is to students, workers, and the public. They should teach about AI: its ownership, limitations, social consequences, and deployment in mass surveillance, violent and repressive policing, and warfare, as well as the environmental costs of its infrastructure. But understanding a technology is not the same as making students dependent on it. Faculty and students deserve a meaningful voice in whether it enters their classrooms, including the right to refuse. A right to refuse means little if opting out carries academic penalties or leaves students without an equivalent route through a course.
Before committing another dollar to AI platforms, universities should weigh those investments against smaller classes, tutoring, disability services, libraries, and the time and resources faculty need to teach. The burden of proof belongs to those proposing to change the conditions of learning. At stake is more than control over software: It is the university's authority over the purposes and human relationships that give education its meaning. Universities should not surrender that authority merely because a technology has become widely available.
