NTU to ditch AI detectors, says technology too unreliable to prove cheating

August 16, 2026 - 04:20

SINGAPORE, Aug 16 - Nanyang Technological University is scrapping its use of institutional AI detection software, admitting the tools are too error-prone to serve as proof of academic misconduct. The decision, confirmed this week, marks a shift in how the university handles the challenge of AI-assisted work in student submissions.

The university said the detectors, which scan essays and assignments for patterns typical of machine-generated text, produced too many false positives. Students who wrote their own work were sometimes flagged as cheaters, while AI-generated passages could slip through undetected. That inconsistency made the software useless as a basis for disciplinary action, according to NTU officials.

Instead of relying on automated flags, NTU will move toward a more holistic review process. Lecturers will compare a student's writing style across multiple assignments, look at draft histories, and hold follow-up interviews when they suspect misuse. The university also plans to update its academic integrity guidelines to clarify when and how students may use AI tools for research, brainstorming, or editing.

The change aligns with a broader trend in higher education. Several universities in the United States and Europe have already backed away from AI detectors, citing similar reliability problems. NTU's move is one of the first by a major Asian institution to publicly acknowledge the limits of the technology.

Students have welcomed the decision, though some faculty members worry about the added workload. The university insists the new approach will be fairer and more transparent, even if it takes more time. NTU will continue to educate students on responsible AI use, but the burden of proof now rests on careful evaluation, not a software score.