From AI Permission to Proof: Procedural Dependence and the Limits of Repair in University Misconduct Decisions

Abstract Evidence that generative artificial intelligence contributed to an assessment does not establish that a student breached its governing rule. This inference gap becomes decisive when universities permit some assistance while prohibiting other forms of substitution. This article develops a conceptual and legal account of procedural continuity in AI-related misconduct decisions, drawing on English and Australian doctrine, assessment scholarship, and procedural justice theory. Its central claim is that safeguards are dependent and conditionally non-compensatory: a later safeguard can repair an earlier defect only if it restores the decision-making function that the defect impaired. The article distinguishes foundational defects, reconstructable participation defects, and irreversible losses of evidence or opportunity. It proposes four conditions for an adequate repair: lawful authority, informational reconstruction, fresh consideration, and an effective response to consequential harm. The analysis also explains why requests for unannounced draft histories can transfer the practical burden of proof to students, and why academic judgment cannot answer every factual or procedural challenge. Short institutional illustrations and two contrasting published complaint summaries expose the mechanisms without serving as an empirical sample. The resulting account shifts university AI governance from the presence of procedural provisions to the preservation, or restoration, of a defensible route from an applicable rule to a proved breach and an effective remedy.