NTU to stop using 'fundamentally unreliable' AI detector from 2027

SINGAPORE: Nanyang Technological University (NTU) will discontinue the use of its artificial intelligence detector from 2027, said the institution in an email to faculty members on Thursday (Aug 13).

The university recognises that the current automated AI detection tools are “fundamentally unreliable”, read the email signed by Dr Tan Seng Chee, the school’s associate vice-provost for education transformation.

These tools also cannot distinguish between unauthorised misconduct and the responsible partnership with AI needed in the curriculum, he wrote.

Since these tools “lack empirical validity”, NTU will deactivate its institutional AI detector at the end of this semester, or Dec 31, said Dr Tan in the email seen by CNA.

AI misuse was a source of controversy for NTU in 2025. In April that year, three NTU students were accused of misusing AI for false and inaccurate citations.

The students – who were told in class that the use of ChatGPT and AI tools was not allowed in the “development or generation” of their essay proposal and long essay – disputed the claims and raised concerns about due process.

All three of them were accused of and penalised for using generative AI in their work. NTU later held consultations with them and upheld the zero mark awarded to one student who appealed the outcome.

NTU will not rely on AI probability scores as evidence of academic misconduct, Dr Tan wrote in his email, noting that this is aligned with emerging best practices across local and international universities.

“Consequently, we are transitioning away from automated surveillance toward assessment designs that evaluate student learning through transparent process-oriented documentation,” he said.

The decision to phase out AI detection tools comes from NTU’s commitment to the well-being of students and faculty, and its intent to preserve a culture of mutual trust, the email read.

For students, relying on “adversarial detection scores” creates an environment of surveillance and distrust, which “undermines the psychological safety” needed for creative exploration and productive struggle, said Dr Tan.

For faculty, AI detection scores introduce administrative noise and cognitive friction, he added.

Reviewing ambiguous probability reports forces teachers into defensive policing roles, leading to burnout from investigating cases and strained relationships between faculty and students, said the associate vice-provost.

“Deactivating these tools frees faculty to focus their time and energy on meaningful pedagogical engagement,” he said in the email.

As of this academic year, all faculty and students have access to NTU’s AI enterprise suite, he said, adding that providing access to institutional AI tools while continuing to run automated AI detection creates mixed signals.

“Transitioning away from automated detection ensures our technological infrastructure aligns fully with our commitment to responsible AI integration, rather than passive surveillance,” said Dr Tan.

According to a guide on how to handle the matter during the ongoing semester, faculty members have been asked to reframe the role of AI detectors.

The guide, which was attached to the email sent out on Thursday, instructed faculty members to avoid describing AI detectors as a threat and that automated scores or flags were not evidence of violations.

The guide also encouraged faculty members to establish standardised disclosures and adapt the sample AI disclosure form for their classes.

CNA has asked NTU what AI detection software the school is using and whether it led to accusations of AI misuse, as well as reached out to other autonomous universities about whether they will be taking similar steps to stop using AI detectors.

Read Full Article At Source