A former Yale Executive MBA student is suing the university in a 13-count federal complaint after he was suspended for one year and given an F for allegedly using generative AI on a final exam. Thierry Rignol paid $208,500 in tuition for the program and says the discipline cost him a valedictorian spot he had earned. The case, filed in February 2025, has now stretched across 125 docket entries and a third amended complaint, with no trial date in sight.
The exam at the center of the dispute was the spring 2024 final for MGT423E, Sourcing and Managing Funds, a four-hour self-timed test that was open-book but closed-internet, with AI tools disallowed. Of 72 students in the course, only Rignol's submission was flagged for possible AI use, in part because of its length and polish. Professor K. Geert Rouwenhorst emailed his dean on June 11 to describe the concerns.
Rignol's professors ran his answers through GPTZero, which returned high probabilities of AI generation on multiple sections. Yale's teaching team also cited overlap between one answer and a response ChatGPT produced to the same question, plus a relatively weak performance on question 5, where AI tools were least useful. Rignol argues the detector was wrong and points to what his complaint calls GPTZero's known bias against non-native English speakers.
Key facts
- 01Rignol paid Yale $208,500 in tuition for its Executive MBA program before being suspended for one year and given an F.
- 02His exam was flagged after GPTZero identified sections as AI-generated; only 1 of 72 students in the course was flagged.
- 03The lawsuit has grown to 13 causes of action across 125 docket entries since it was filed in February 2025.
- 04Rignol says he wrote the 4-hour open-book exam in Apple Pages, not Microsoft Word, which Yale had repeatedly requested.
- 05As a demonstration, Rignol ran works by a Yale dean and former president through GPTZero, which returned a 100% AI probability.
The evidentiary weakness of AI detectors is not a fringe view. Even Yale, in the university's own guidance, has acknowledged that policing AI use through detection tools is infeasible. Rignol dramatized the point at his November 8 Honor Committee hearing by submitting GPTZero scans of works by a Yale dean and a former Yale president — some published more than 30 years ago — that the tool rated as 100% AI-generated.
The messier part of the case is not the detector but a file-format dispute. Throughout summer 2024, finance professor James Choi, who led the Honor Committee, repeatedly asked Rignol to produce the Microsoft Word file used to draft his exam. The requests came on August 10, August 12, August 16, and August 19. No file was provided.
Choi's August 10 email carried an explicit warning about the stakes of non-cooperation with the committee. Rignol's lawsuit argues he was being pressured to falsely confess under threat of expulsion, and alleges that one dean invoked the possibility of deportation. Yale counters that Rignol was in the United States on an investor visa, not a student visa, and that discipline would not have triggered visa revocation.
“Through Dean Wendy Tsung, I have twice asked you to send us the Microsoft Word file that produced [the] PDF file you submitted for the exam. I am now asking you directly.”— James Choi, Yale finance professor and Honor Committee chair
At the November 8 hearing, Rignol disclosed for the first time that he had not used Microsoft Word to write the exam. He had used Apple Pages. Yale's filings express incredulity that this detail was withheld across months of correspondence in which Choi specifically and repeatedly asked for a Word file. Rignol's position, as summarized by Yale's lawyers, was that he had no obligation to correct the professor's assumption about which program he used.
Rignol produced the Pages file shortly after the hearing ended around 2 pm that day. Assistant Dean Wendy Tsung then asked him to return to campus by 5:30 pm so the committee could examine his laptop. Rignol, who had left for the day, declined and asked to reschedule for the following week. That evening at 8:28 pm, Choi sent the letter that led to the suspension.
The 13 causes of action now include breach of contract, civil rights violations, defamation, invasion of privacy, emotional distress, and unfair trade practices. Rignol also alleges the process was designed to punish what he describes as protected conservative political speech advocating smaller government, pro-business policies, and skepticism of DEI. He is seeking damages without limitation and asks the court to reverse the F and expunge his disciplinary record, even though Yale says he has already completed the suspension and graduated.
Yale, for its part, portrays Rignol as evasive throughout the investigation — stalling on file production, questioning the composition of the Honor Committee days before the hearing, and volunteering the Apple Pages detail only after months of specific requests for Word documents. Whether the underlying exam was AI-assisted may never be definitively resolved, because the artifacts that could have shown edit history were not produced when they would have mattered most.
The case is one of the first federal lawsuits to sit squarely on the question of how universities can, or cannot, use AI detectors in academic discipline. GPTZero and similar tools carry documented false-positive rates and known biases against non-native English writers and highly formal prose, which makes them poor sole evidence in a hearing where the stakes are suspension or expulsion. Schools that build discipline cases primarily on detector output are inviting exactly this kind of protracted litigation.
For AI vendors and universities alike, the Yale case sets a template worth watching. Detectors will keep getting used because instructors have no better tool, but the legal exposure of relying on them — especially without corroborating evidence like edit history, keystroke logs, or draft files — is now on display in a docket that has already produced 125 entries. Expect universities to shift toward process-based evidence, and expect vendors selling AI-detection services into education to face harder questions about their false-positive rates in court, not just in marketing copy.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
