For four years the campus argument about artificial intelligence has run in one direction: institutions trying to stop students from using it. That argument has now reversed. The New York Times reported this week that students at a growing number of universities are pushing back against mandates that require AI use in coursework, and are demanding opt-outs from assignments where the tool is compulsory.
The objections are not only ethical, though environmental cost, copyright and distrust of the vendors all feature. They are educational. Students argue that being made to hand a task to a model removes the practice the assignment exists to provide, and that a degree built on delegated thinking is worth less than the one they were sold.
Institutions are beginning to codify the middle ground. On 12 November, NYU's Tandon School of Engineering released Protecting Learning, a guide that sorts every assignment into one of three categories: red, where AI is prohibited; yellow, where limited assistance is permitted; and green, where AI use is part of the assessed work. Instructors set a course baseline and can move individual assignments between categories. Where independent competence is the point, the guide recommends AI-free checks such as oral questioning, supervised quizzes or in-class exams.
The guide also takes a position on detection that most universities have avoided: a detector score alone cannot establish a violation, and instructors should gather multiple forms of evidence and speak to the student before reaching a finding. It is not yet official Tandon policy, with public comment open through November and a first revision planned for January 2027.
Elsewhere the split runs the other way. The University of Chicago will require every first-year student to take a writing course in which AI is not allowed, starting next fall. At Eastern Illinois University, students are asking for limits not on themselves but on their professors, seeking rules on where AI may be used in grading, feedback and course materials.
The data explains the tension. HEPI research puts the share of UK students using generative AI in assessed work at 94%. A study by the Office for Students and Advance HE, drawing on nearly 100 students and more than 200 staff, found students who felt pressured to use AI to save time or because it looked like the reliable route to a good grade, and who worried about what they called cognitive offloading, the outsourcing of engagement with course material. Most of them still believed AI literacy would be essential after graduation.
Employers are not reassured. A University of Alabama study published on 12 November, built on 56 sources, interviews with 13 industry leaders and a survey of 110 mostly US employers, rated human oversight as the single most important capability, and judged recent graduates least prepared on ethics and policy and on verifying AI output. A graduate who can produce a report with a model but cannot tell whether it is accurate is the specific failure mode being described.
That is why the opt-out request is harder to dismiss than it first appears. Universities spent years building detection regimes, and those regimes are losing: rewriting tools and typing simulators are sold openly, and some vendors sell detection to institutions and evasion to students. Against that backdrop, requiring students to use the technology inverts the burden of proof. The defensible question is the one NYU Tandon poses, assignment by assignment: is AI use itself a learning outcome, or merely one available tool? If a student can demonstrate the outcome without it, mandating it serves no pedagogical purpose, and if AI literacy is the outcome, refusing to use it is like refusing to drive during a driving test.
Expect syllabi to start declaring AI status per assignment rather than per institution, and expect procurement to follow, as students ask why their tuition funds tools they are told they must use. The schools that formalise opt-outs first will set the template the rest copy.
Comments (0)
Log in to join the discussion
Log InNo comments yet