How a Yale AI-cheating dispute became a 13-count federal lawsuit - Ars Technica

Skip to content

AI

Biz & IT

Cars

Culture

Gaming

Health

Policy

Science

Security

Space

Tech

Forum

Subscribe

Story text

Size

Small<br>Standard<br>Large

Width

Standard<br>Wide

Links

Standard<br>Orange

* Subscribers only

Learn more

Pin to story

Theme

Search

Sign In

Sign in dialog...

Text<br>settings

Story text

Size

Small<br>Standard<br>Large

Width

Standard<br>Wide

Links

Standard<br>Orange

* Subscribers only

Learn more

Minimize to nav

Thierry Rignol is furious at Yale.

Rignol paid Yale $208,500 in tuition for its Executive MBA program. But after he was accused of cheating, the school suspended him for a year and gave him an F in the course Sourcing and Managing Funds.

This happened despite the fact that Rignol was a “top student, on track to graduate first in his class,” he says. As a result, Rignol missed out on being named class valedictorian, an honor he claims to have earned “under Yale’s own stated criterion.”

So Rignol sued the school. His lawsuit has been running since February 2025. It opened with a motion to “Proceed in Fictitious Name”—which was denied—and has since sprawled across 125 docket entries and a highly unusual third amended complaint. It appears to be nowhere near trial.

What issue was big and loud enough to produce all this sound and fury? Artificial intelligence, of course.

Rignol’s final exam in Sourcing and Managing Funds was flagged as possibly composed with the aid of generative AI. Rignol’s professor used an AI detection tool called GPTZero on Rignol’s text. The tool said that whole sections were likely AI-generated. But such tools have well-known weaknesses; most are unreliable, and even Yale has noted that trying to police AI use through detection tools is infeasible.

In Rignol’s telling, his long, perfectly formatted test answers were the product of his own intelligence. “It was entirely expected,” his lawsuit says, “that [Rignol’s] exam writing would be thorough, well-organized, and display near perfect punctuation and grammar, attributes consistent with his academic excellence.”

There was also the matter of his French background. Rignol calls out “GPTZero’s known bias against non-native English speakers” like himself, and he argues that detectors can mistake the formal, structured prose of such writers for AI-generated text.

But it’s not just tools like GPTZero that enrage Rignol. He believes that Yale itself is against him and that the whole “sham” disciplinary process was designed to censor his “protected conservative political speech” after he advocated in courses for “smaller government, pro-business policies, and skepticism of DEI.”

After several expansions, Rignol’s case now contains 13 separate causes of action against Yale, including breach of contract, civil rights violations, emotional distress, unfair trade practices, defamation, and invasion of privacy. He wants damages “without limitation” to cover his physical well-being, emotional state, reputation, “past and future economic losses,” and damage to career prospects.

Despite everything, Rignol’s lawsuit still demands that Yale take him back (though the university says he has already returned from his suspension and has graduated). Rignol also wants his F reversed and his disciplinary records expunged.

But was Rignol actually the victim of a bad detection algorithm and political hostility at Yale? The school doesn’t think so—and it wants to know why Rignol took months to turn over a key file.

Incomplete

The saga began with the final exam in the spring 2024 session of MGT423E, Sourcing and Managing Funds. The test ran for four hours and was self-timed; it was “open book” but “closed Internet.” AI tools were strictly disallowed. Students wrote answers on their computers, generated PDFs from those files, and submitted the PDFs to their instructors. Of 72 students, only Rignol’s exam was flagged by a teaching assistant for possible AI use, in part because of its unusual length.

One of the course’s professors, K. Geert Rouwenhorst, emailed his dean on June 11 to describe the suspicions about the exam. While Rignol had argued largely over the use of tools like GPTZero, Rouwenhorst made clear that his teaching team had multiple reasons for concern.

First, he said, GPTZero thought it likely that AI was used in several answers. Second, one answer showed “substantial overlap” with output generated by ChatGPT in response to the same question. Third, Rignol “performed relatively poorly on question 5, where AI tools were the least helpful.” Finally, the teaching team questioned—though without testing the theory—whether anyone could have produced an exam of that length and polish within four hours.

The next day, Rignol learned that he would receive an “Incomplete” in the course while the issue was investigated.

Over the summer of 2024, various administrators looked into the matter. Rignol claimed innocence, saying the exam...