Generative artificial intelligence has made refusal newly visible in higher education. Students, academics, researchers, professional staff and institutional leaders may reject, restrict, contest or delay AI use for reasons grounded in authorship, learning authenticity, privacy, environmental cost, labour politics, epistemic mistrust, professional identity or educational integrity. This Comment develops AI refusal as a conceptual and diagnostic category rather than a single measurable behaviour. It argues that refusal in higher education should be understood as a spectrum of non-use, opt-out, restriction, bounded use and institutional delay, whose meaning changes across roles because universities are role-structured and publicly accountable institutions. Recent evidence suggests that outright student non-use is relatively low in some jurisdictions, while faculty opt-out, institutional hesitation and assessment uncertainty remain more substantial and consequential. The article distinguishes the right to refuse from the duty to understand, arguing that private refusal should be protected, while role-based refusal must be accountable where AI materially shapes learning, teaching, assessment, research, administration or governance. It also recognises that refusal is not always available as an individual act when AI is embedded in institutional infrastructures. Refusal is proposed as diagnostic evidence of what higher education believes is at risk, and how risks should be governed.
Jason Zagami (Tue,) studied this question.