In April 2023, the plagiarism-detection company Turnitin switched on an AI-writing detector for its customers with less than 24 hours’ notice and, initially, no way to turn it off. Four months later, on August 16, 2023, Vanderbilt University’s Brightspace team published a detailed statement explaining why it was disabling the feature.
The central concern was false positives. Turnitin advertised a 1 percent false-positive rate, but Vanderbilt did the arithmetic: it had submitted about 75,000 papers to Turnitin in 2022, so a 1 percent error rate would have wrongly flagged roughly 750 student papers as AI-written - each one a potential, baseless academic-integrity case. The university also cited a lack of transparency about how the detector worked, evidence that such tools disproportionately flag writing by non-native English speakers, and a broader judgment that AI detection is a fundamentally hard - perhaps impossible - problem that will only get harder as AI writing improves. Its conclusion was blunt: “we do not believe that AI detection software is an effective tool that should be used.”
Why business readers should care: this is a textbook case of why a low advertised error rate can still be unacceptable at scale. A 1 percent rate sounds excellent until it is multiplied by 75,000 cases and the cost of each error is a false accusation against a student. Base rates and consequences, not headline accuracy, decide whether an AI classifier is safe to deploy.