The running list
Universities stepping back from AI detectors.
Detection is quietly being switched off. These institutions have disabled it, discontinued it, cancelled the contract, or declined to turn it on in the first place — each in their own published words, linked below.
8 institutions · last updated August 19, 2026
Washington State University
United StatesContract cancelledFebruary 2026The Provost’s Office cancelled the Turnitin AI-detection contract, citing false positives and the distress they caused students; Turnitin’s plagiarism (text-matching) package stays available in Canvas.
Instead: Text-matching only; faculty pointed to SafeAssign for similarity checking
Curtin University
AustraliaDisabledJanuary 1, 2026Turnitin’s AI-writing-detection feature was switched off across all campuses and study periods; text-matching originality checking continues unchanged.
University of Waterloo
CanadaDiscontinuedSeptember 2025AI-detection functionality in Turnitin was withdrawn from all users after internal testing flagged wholly human-written text as AI-generated.
Instead: Assessment redesign and AI literacy, over detection
Vanderbilt University
United StatesDisabledAugust 16, 2023Turnitin’s AI detector was disabled university-wide after months of testing, citing false positives, bias against non-native English writers, and no visibility into how the tool decides.
Johns Hopkins University
United StatesNot recommendedThe Center for Teaching Excellence and Innovation advises against AI-detection tools: they return a vague probability rather than evidence an instructor could show a student.
Instead: Clear per-course expectations for AI use, set up front
University of California, Berkeley
United StatesNever enabledA Fall 2023 pilot of Turnitin’s AI detection returned inconsistent results, so the tool was not enabled campuswide; instructors are advised not to rely on AI detectors.
Instead: Conversation with students about what is and isn’t allowed in the course
University of California, Los Angeles
United StatesNever enabledUCLA opted out of Turnitin’s AI-writing-detection preview feature over unanswered questions about it, a call shared across much of the UC system.
Instead: Exploring writing-process insight tools rather than detection
Yale University
United StatesNever enabledThe Poorvu Center does not enable Turnitin’s AI-detection feature in Canvas and does not endorse AI-detection software, calling the tools unsuitable for high-stakes decisions.
Instead: Talking with the student about the work itself
How this list is maintained
A row only goes in when the institution has said so itself, on its own site. News round-ups and third-party trackers list plenty more schools than are here; we don’t add one on that basis, because a table about honest signals is worthless the moment it overstates one. Where a source gives no date, the date column stays empty rather than being guessed.
These are also different acts, and the label says which: a campus that disabled a feature it had running is not the same as one that never enabled it. None of these institutions use, endorse, or have any relationship with Folio — they’re here because they published a decision, nothing more. Policies change; the linked page is always the authority, not this one.
Know of an institution that belongs here, or one whose policy has changed? Send us the link and we’ll check it against the same bar.
Why it keeps happening
Turning the detector off doesn’t answer the question it was asked.
The reasons repeat across every row: a score is a probability, not evidence; the false positives fall hardest on non-native English writers; and an integrity board can’t act on a number nobody can reproduce. But an institution that switches detection off still needs to know who wrote the work.
That’s the case for provenance instead of prediction: a record of how the work actually took shape, captured as it’s written and independently verifiable at submission — no guessing, and no watching how a student types.
Reassessing your detection stack?
We’ll walk your integrity office through what proof-of-process actually looks like — no sales pitch, just a straight answer.
Contact us