A closed-internet, no-AI-tools MBA final. A long, polished submission. Then an algorithm decides the prose sounds too good to be genuine. That’s roughly what happened to Yale Executive MBA student Thierry Rignol in spring 2024, and the fallout has grown into a 13-count federal lawsuit with more than 125 docket entries. What started as an academic integrity investigation has become a stress test for every university leaning on AI-detection software as evidence.
When the Algorithm Flags You
GPTZero output, writing style, and a delayed file disclosure turned one exam dispute into a procedural quagmire.
The course was “Sourcing and Managing Funds” at Yale’s School of Management — open-book, but no internet, no AI tools permitted. Yale professors flagged Rignol’s submission after GPTZero indicated portions were likely AI-generated. They also cited its unusual length, polish, and overlap with ChatGPT-style output. Rignol pushed back: his formal writing, academic strength, and non-native English background are exactly the kind of inputs that trigger false positives on these tools.
AI detectors have well-documented reliability problems. That fact alone should give any institution pause before treating their output as evidence.
Then came the file-format twist. Yale repeatedly asked Rignol for the underlying document used to create his submitted PDF. Months passed. Eventually he disclosed it was an Apple Pages file — not Microsoft Word. The federal judge reportedly noted the obvious question: why wouldn’t a reasonable person immediately clarify that? Yale’s Honor Committee found him liable both for the original allegation and for “not being forthcoming.” The penalty: a one-year suspension and an F in the course.
One Suspension, 13 Claims
The lawsuit has ballooned far beyond academic integrity into breach of contract, civil rights, and defamation territory.
Rignol filed suit in federal court in February 2025. The complaint has since expanded to 13 causes of action:
- Breach of contract
- Civil-rights violations
- Emotional distress
- Unfair trade practices
- Defamation
- Invasion of privacy
A federal judge has already denied at least one request for immediate reinstatement. Yale maintains its process was justified and that Rignol’s months-long delay in producing the source file was the core issue.
This case is the canary for every school quietly treating AI-detection output like forensic evidence. Relying on GPTZero as proof of cheating resembles using a breathalyzer with a known margin of error in a criminal trial — the confidence of the reading outpaces its actual reliability. Formal prose, structured arguments, and non-native English patterns all raise flags that have nothing to do with ChatGPT.
Whether Rignol cheated remains legally unresolved. The harder question is whether AI-detection software can carry the weight of a disciplinary finding that derails someone’s career. Universities betting on these tools may soon have to answer that in court.
