Yale faces a 13-count federal lawsuit over AI cheating detection

Yale Faces a 13-Count Federal Lawsuit Over AI Cheating Detection

Thierry Rignol paid Yale $208,500 in tuition for an Executive MBA. What he got instead was an F, a one-year suspension, and a federal lawsuit running since February 2025 that has sprawled across 125 docket entries with no trial in sight 1.

The dispute centers on a four-hour, open-book final exam in a course called Sourcing and Managing Funds. AI tools were prohibited. Of 72 students in the cohort, only Rignol's exam was flagged for suspected AI use, initially by a teaching assistant who noted its unusual length 1.

A professor then ran Rignol's answers through GPTZero, an AI detection tool that flagged whole sections as likely AI-generated 1. The result set in motion a legal battle that is becoming a landmark test of whether universities face legal liability when they build disciplinary systems on detection tools that researchers say are unreliable.

The bias dimension

Rignol is a French non-native English speaker. His lawsuit argues that GPTZero has a known bias against writers like him, citing research showing that AI detectors routinely flag formal, structured prose from non-native writers as machine-generated 1.

Independent research backs that claim. A 2023 study from researchers at Stanford University found that GPT detectors consistently misclassify non-native English writing as AI-generated, while accurately identifying native English writing 2. The study cautioned against deploying these detectors in evaluative or educational settings, warning they may penalize writers with limited linguistic range 2.

This is what converts the case from a single student's grievance into a systemic risk story. Universities deploy AI detection tools as if they were breathalyzers: binary, reliable, standards-based instruments that produce a clean verdict. The evidence suggests they function closer to polygraphs: directional, biased, and legally contested. The only exam flagged in a cohort of 72 belonged to the non-native English speaker. That is the exact pattern Stanford researchers predicted.

Yale's counter-narrative

Yale's filings tell a different story. Professor K. Geert Rouwenhorst outlined multiple reasons for suspicion beyond GPTZero when he emailed his dean on June 11, 2024. One of Rignol's answers significantly overlapped with ChatGPT's response to the same question. Rignol also scored lower on the exam question where AI tools would have been least useful. The teaching team questioned, without testing the theory, whether anyone could produce an exam of that length and polish within four hours 1.

Finance professor James Choi, who led the Honor Committee investigation, repeatedly asked Rignol for the underlying file used to draft the exam. Yale says that file could have shed light on whether Rignol used AI tools. Choi emailed Rignol at least four times between August 10 and August 19, 2024. No file arrived 1.

The stakes beyond Yale

Rignol's lawsuit now contains 13 separate causes of action, including breach of contract, civil rights violations, emotional distress, unfair trade practices, defamation, and invasion of privacy 1. He wants damages without limitation, his F reversed, and his disciplinary records expunged 1.

The case matters beyond New Haven. If discovery forces Yale or GPTZero to disclose the tool's internal accuracy data, the precedent could reach every campus using AI detection. Even Yale itself has acknowledged the problem: the university has noted that trying to police AI use through detection tools is infeasible 1.

Rignol v. Yale is asking a federal court to decide what happens when an institution builds its accountability system on a tool its own research community says does not work. The answer will shape how universities police academic integrity across the country.