A professor at Brown University says at least 50 students cheated on a midterm exam using ChatGPT, and he’s calling it the biggest known academic integrity scandal in the entire Ivy League. According to Hacker News, the professor is Roberto Serrano, the Harrison S. Kravis University Professor of Economics, who caught the fraud in ECON 1170, an advanced undergraduate course in mathematical economics. The midterm was administered on March 5, and what happened next gives us a rare, documented look at AI cheating at scale.
This matters because it’s not a hunch or a vibe check. Serrano has receipts.
What Happened
Servrano switched his exams this year to take-home, closed-book format, a style with some tradition at Ivy League schools. The idea is generous: give students near-unlimited time, then make the questions harder to see how far they can push. He changed some of the model assumptions from class and asked students to prove whether certain statements held true under the new rules.
The results were suspicious from the start:
- Enrollment jumped to 86 students. Serrano had never had more than 30, and some years just eight.
- The midterm average came in at 96 out of 100. Forty students scored a perfect 100.
- Graders flagged “unusual passages that coincided with results obtained after running the questions through ChatGPT,” according to Hacker News.
Servrano didn’t void the midterm. Instead he set a trap. He told students the final, worth 50% of the grade, would be in-person, and warned that if the grade distribution didn’t match the midterm, only the final would count.
The Numbers Told the Story
The final exam average dropped to 48 out of 100. Half the score of the midterm. Of the 89 students who took the midterm, only 59 showed up for the final. And here’s the detail that seals it: of the 27 who skipped the final, 22 had scored a perfect 100 on the midterm.
“The empirical evidence of fraud is overwhelming,” Serrano said. Hard to argue with a game theorist who built a test specifically to surface the cheaters. This is a clean natural experiment, and the students who relied on AI essentially self-selected out.
The Part That Should Worry Educators
What stands out here isn’t the cheating. It’s the institutional response. When Serrano reported the case to Brown’s leadership, he says the president responded with “absolute silence.” The dean stayed quiet too until Serrano escalated to the Academic Code Committee, which acknowledged the incident as “a wake-up call.”
Servrano, who has taught at Brown for 34 years, thinks that’s nowhere near enough. “Academic integrity is a value worth defending,” he said. “The faculty cannot be left on its own in a battle that is decisive if we want to preserve the future of higher education.” His fix: “We need to publicly admit the seriousness of the situation and open up a broad debate about the real extent of the problem.”
Why This Is Bigger Than One Classroom
Take-home and unproctored assessments have been a quiet standard at elite schools for decades. They rewarded trust. Generative AI breaks that model in a way plagiarism checkers never quite did, because the output is original each time and tuned to the exact prompt. Detection tools are unreliable, so faculty are left reverse-engineering behavior, which is exactly what Serrano did.
For next year, he’s already changing course:
- Weekly exercises will no longer count toward the final grade, since they can be done with AI.
- No more take-home exams, no matter how pedagogically useful they are.
That second point is the real cost. A genuinely good teaching tool is getting retired because it can’t survive contact with ChatGPT. Expect more of this across higher ed: a quiet rollback to in-person, proctored, handwritten assessment, plus louder fights over whether administrators will back the faculty catching the fraud or just absorb the embarrassment.
Servrano forced the issue with evidence instead of suspicion. The open question is whether universities treat cases like his as isolated incidents or as the warning they clearly are. More details are available at the original source.