University Wrongly Accused Students of Using AI: Case Study from Australia

What happened

• At Australian Catholic University (ACU), a paramedic student (called “Madeleine” in reports) was accused of using AI to cheat on an assignment.  

• The accusation came via an “Academic Integrity Concern” email. The student was asked to explain why the university believed she used AI. While under investigation, her transcript was marked “results withheld,” which affected her ability to apply for graduate positions.  

• She says she had done nothing wrong, provided evidence (handwritten and typed notes, etc.), but the university’s case was based largely on a report from Turnitin’s AI-detection tool that flagged much of her text (84%) as “supposedly written by AI.”  

• ACU recorded nearly 6,000 referrals for alleged academic misconduct in 2024; about 90% related to AI use. But many of those accused had done nothing wrong. Roughly one quarter of AI-use referrals were dismissed after investigation, and any case where Turnitin’s AI detector was the sole evidence was “dismissed immediately.”  

• The student took six months for the process, during which the accusations and transcript status caused real harm (missed opportunities). She was eventually cleared.  

Why this is problematic

• Reliance on imperfect detection tools: Turnitin (and similar tools) can issue false positives. They warn that AI detection reports may not always be accurate and should not be used as the only evidence. But in many cases they are being used as the main or sole basis for accusations.  

• Burden of proof on students: Once accused, students must prove their innocence, often gathering drafts, timestamps, notes, or search histories, which is stressful and sometimes invasive.  

• Delays and impact: The investigation processes are often slow; meanwhile students suffer academic, financial, and emotional harm. Delayed graduation, withheld results, or damaged reputation are common.  

• Ambiguity and inconsistency: Policies around what counts as “unauthorised AI use” vary; staff may differ in what they expect or how strictly they apply rules. Also, detection tools often flag clean writing, paragraphs or formatting unusual for the student, even when that can be explained.  

What ACU has done/is doing

• ACU decided to stop using Turnitin’s AI detection tool as of March 2025 after finding it ineffective.  

• They acknowledge that their investigations were not always timely and say improvements have been made.  

• They’ve introduced new modules on ethical use of AI for staff and students.  

Broader Implications

• As AI tools become more available, universities are under pressure to police academic integrity. But detection is hard. False detections can do a lot of damage.

• There’s growing debate over whether banning or heavily policing AI use is helpful: many argue education (teaching how to use AI appropriately) and redesigning assessments might be better.  

• Also, the fairness of using automated tools without sufficient transparency or human oversight is under scrutiny.