A Yale Executive MBA student who paid $208,500 in tuition is suing the school in federal court after being suspended for one year and given an F on a final exam flagged as AI-generated by the detection tool GPTZero.
The lawsuit, filed in February 2025 and now containing 13 counts including breach of contract and civil rights violations, hinges on whether the detection tool was reliable and whether the student's delay in producing his underlying Apple Pages file (which professors had asked for weeks before he clarified he did not use Microsoft Word) constitutes dishonesty or justified non-cooperation.
The case highlights the unreliability of AI detection tools and the challenge universities face in proving AI use without transparent methods.
What happened
Thierry Rignol, a top student in Yale's Executive MBA program who paid $208,500 in tuition, was suspended for one year and given an F after his exam was flagged by GPTZero, an AI detection tool, as possibly AI-generated. He filed a federal lawsuit in February 2025 now containing 13 causes of action, including breach of contract, civil rights violations, and defamation. The core dispute centers on whether Rignol used generative AI on a four-hour final exam in Sourcing and Managing Funds, and on his months-long delay in providing the underlying Apple Pages file that generated his PDF submission.
Why it matters
The case exposes the unreliability of AI detection tools—Yale's own Honor Committee found GPTZero flagged works by a Yale dean and former president (some over 30 years old) as '100% probability' AI-generated, clearly false results. Rignol also argues the tools are biased against non-native English speakers like himself, mistaking formal, structured prose for AI text. The lawsuit raises questions about how universities should handle suspected AI use without proven, transparent detection methods, and whether academic discipline based on tool output alone can withstand legal scrutiny.
What to watch
The case has accumulated 125 docket entries and remains nowhere near trial. Yale contests Rignol's account, noting he did not disclose he used Apple Pages (not Microsoft Word) until the November hearing, weeks after being asked for the file, and then provided it only hours before the Honor Committee demanded to examine his laptop. A federal judge expressed skepticism about Rignol's conduct during a hearing, suggesting a reasonable professional would have immediately clarified the word processor discrepancy.
Thierry Rignol, a student in Yale's Executive MBA for Executives program, paid $208,500 in tuition. He was, by his account and reportedly by Yale's own record, a top student on track to graduate first in his class. But in the spring 2024 session of MGT423E, Sourcing and Managing Funds, his final exam—a four-hour, self-timed, open-book but closed-Internet test written on his computer—was flagged as possibly AI-generated. Of 72 students, only Rignol's submission was flagged by a teaching assistant, in part because of its unusual length. Professor K. Geert Rouwenhorst emailed his dean on June 11 with multiple concerns: GPTZero flagged several answers as likely AI-generated; one answer showed "substantial overlap" with ChatGPT output for the same question; Rignol performed poorly on question 5, where AI tools were least helpful; and the teaching team doubted anyone could produce such a polished exam within four hours. On June 12, Rignol was told he would receive an "Incomplete" while the matter was investigated. Over the summer, finance professor James Choi, leading the Honor Committee investigation, asked Rignol repeatedly for the underlying file used to create his PDF submission—first via email through Dean Wendy Tsung, then directly. On August 10, Choi wrote: "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. I remind you that when a student has been found by the Honor Committee to have lied to or not been forthcoming with the Committee, the most common punishment is permanent expulsion from Yale." Choi followed up on August 12, August 16, and August 19. No file arrived. In September, the Honor Committee sent an official notification accusing Rignol of "improperly utiliz[ing] AI on the final exam." Rignol asked to postpone the hearing until after his wedding in Europe; the request was granted. In early November, back from Europe, Rignol challenged the committee's composition. The hearing proceeded on November 8. At the hearing, Rignol maintained his innocence and submitted GPTZero scans of works by a Yale dean and a former president, demonstrating that the tool had flagged these—some published over 30 years ago—as "100% probability" AI-generated, which he called clearly false. The Honor Committee again asked for the underlying Word document. This time, Rignol explained he had written the exam in Apple Pages, not Microsoft Word, and had no relevant Word files. Around 2 pm, the hearing ended. At approximately 3 pm, Rignol provided the Apple Pages file. At 3:30 pm and again at 4 pm via text, Dean Tsung requested he return to campus by 5:30 pm so the committee could examine his laptop. At 5:07 pm, Rignol texted that he could not meet that deadline and suggested the following week. He did not return. At 8:28 pm the same evening, Rignol received a letter from Professor Choi. The Honor Committee had not ruled on whether Rignol used AI, but it found him liable for "not being forthcoming"—a separate honor code violation—without having provided him advance notice or opportunity to defend himself against that specific charge. Rignol was suspended for one year. In late November, the Honor Committee further decided he had "violated the rules of the Sourcing and Managing Funds final exam" and imposed an F. Rignol appealed both the suspension and the grade to Yale administrators; both appeals were denied. In February 2025, he sued Yale in federal court. His complaint now spans 13 causes of action: breach of contract, civil rights violations (including claims that the discipline was retaliation for his "protected conservative political speech" advocating smaller government, pro-business policies, and skepticism of DEI), intentional and negligent infliction of emotional distress, unfair trade practices, defamation, and invasion of privacy. He seeks damages "without limitation" for physical well-being, emotional state, reputation, "past and future economic losses," and damage to career prospects, and demands Yale reverse his F, expunge his disciplinary records, and reinstate him (though Yale notes he has already returned from suspension and graduated). Yale disputes his account, pointing to his months-long silence about using Apple Pages rather than Word, his failure to volunteer the correction after weeks of requests, and his refusal to return to campus the same evening to allow the committee to examine his laptop. A federal judge expressed skepticism at a hearing, asking: "Wouldn't a reasonable professional person who was trying to be cooperative with a proceeding upon getting not just one email but many emails asking for the underlying document that was used to create a PDF say, 'Oh, I didn't use Word. I used Pages, a different word processing [program]?'" Two years after the contested exam, the case continues with 125 docket entries accumulated and no trial date in sight.
The Rignol case sits at an intersection of three persistent problems in higher education's response to generative AI: the unreliability of detection tools, the opacity of investigation processes, and the difficulty of establishing guilt without them. GPTZero's flagging of Rignol's exam triggered the entire disciplinary chain, yet during the November hearing, Rignol demonstrated that the same tool had scored 30-year-old published works by Yale faculty as "100% probability" AI-generated—a clearly false result that undermined the committee's reliance on the tool. Yale acknowledged in its own argument that detection tools are infeasible for policing AI use, yet used one as the initial red flag. Rignol's broader defense—that his formal, well-organized prose reflected both his exceptional academic standing and his background as a non-native English speaker—raises a genuine concern: detection tools may systematically penalize writers whose English is learned and structured rather than conversational. Yale's response pivots away from the tool's reliability and instead emphasizes Rignol's conduct: his failure to clarify, weeks into questioning, that he had used Apple Pages rather than Word, and his refusal to return to campus the same evening to allow the committee to examine his laptop. From Yale's perspective, Rignol's opacity about the file itself—not the file's content—violated the Honor Code's requirement of candor. From Rignol's, Yale moved the goalposts from asking for "Word" files to demanding laptop access without advance notice. The federal lawsuit's sprawl—13 counts ranging from breach of contract to defamation—reflects the absence of a clear legal path: detection-tool evidence alone is too weak to survive scrutiny, but investigation procedures and due-process timing offer sharper attack surfaces. With 125 docket entries accumulated and no trial date in sight, the case may resolve not on whether Rignol actually used AI, but on whether Yale followed its own stated procedures fairly.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
The AI news that matters, in one minute each morning.
Sign up free