Brown University AI Cheating: 50 Students, Near-Perfect Scores Exposed

PROVIDENCE — Brown University AI cheating has become a serious conversation at one of America’s most elite campuses after economics professor Roberto Serrano said he holds compelling evidence that at least 50 students cheated on a midterm in an advanced economics course on March 5. One question surfaced immediately: can universities still protect academic integrity when AI can answer exam questions in seconds?

Serrano, the Harrison S. Kravis University Professor of Economics at Brown, said the cheating occurred in ECON 1170 — a notoriously demanding undergraduate mathematical economics course. He described his evidence as “convincing” and called the incident the largest cheating scandal he has encountered at Brown, and possibly the largest known case across the Ivy League.

“The empirical evidence of cheating is very strong,” Serrano said in a phone interview from Providence, Rhode Island, as reported by El País. Short. Definitive. He also urged Brown to move beyond quiet internal warnings and open a public debate about AI’s impact on higher education.

The numbers that put the whole campus on alert

What made this case stand out was not just the number of students allegedly involved — it was the results themselves. Of 89 students who sat the midterm, the class average hit 96 out of 100. Forty students scored a perfect 100.

Investigators then found patterns they deemed abnormal. Several answers contained sections that, according to Serrano, matched outputs produced when the same questions were fed into ChatGPT. That is when suspicion of AI-assisted cheating solidified into something harder to dismiss. Not a hunch. A traceable pattern.

Serrano chose not to cancel the midterm. Instead, he warned students that the final exam would be held in person and would carry decisive weight if the grade distribution looked implausible compared to midterm scores. The outcome was stark. The final exam average collapsed to 48 out of 100.

What made the picture even clearer: of the 27 students absent from the final, 22 had previously scored a perfect 100 on the midterm. That alignment was difficult to explain away.

A university response Serrano found too muted

Serrano said he reported the situation to senior Brown University officials, but found the institutional response cold. The university president, he said, stayed silent. The dean offered no comment until the case reached the Academic Code Committee. Only then did the university describe the incident as a “wake-up call.”

For Serrano, that was not enough. He argued the university needed to be candid about the scale of the threat AI poses to traditional examinations. “Academic integrity is a value worth defending. Faculty should not be left to fight this defining battle alone if we want to preserve the future of higher education,” he said.

That statement hit a nerve. Many instructors across campuses face similar problems without adequate tools. AI can polish answers, construct arguments, and even replicate a student’s writing style. In courses assessed through essays or take-home assignments, the line between assistance and cheating keeps getting thinner.

Why take-home exams became the weak point

This year, Serrano had chosen a closed take-home format — common at some Ivy League institutions. Students get more time to complete questions, which is supposed to justify harder problems and deeper thinking. The goal: measure real understanding.

The problem is that the same format creates more space for AI. No one is watching. Students can open a laptop, generate polished answers with machine help, and submit work that looks entirely academic. From the outside, everything appears legitimate. From the inside, not necessarily.

That is why Serrano decided to overhaul his assessment approach for the next academic year. Weekly assignments will no longer count toward the final grade. Take-home exams are gone entirely. No exceptions.

It may sound harsh. But for a professor who has spent 34 years teaching at Brown, the alternative carries a steeper cost. If universities allow AI to erode the honesty of assessments, grades lose their meaning. And when grades can no longer be trusted, degrees lose theirs too.

From mathematical economics to a debate about the future of college

Serrano is not a peripheral figure. He is recognized as a leading scholar in the application of game theory to market analysis, holds a doctorate from Harvard, and chose Brown specifically to focus on research and teaching. In 2024, he received the King of Spain Prize for Economics.

That track record gives his words weight. He is not a professor in a panic over a new technology. He is an academic who understands exactly how incentives work. In game theory terms, people choose the most efficient path when oversight is loose and consequences are small. AI, applied to exams, offers precisely that incentive structure.

Serrano’s own story also matters here. He lost his sight at 17 due to retinal dystrophy, learned Braille, completed his education, and built an academic career from Harvard to Brown. Today, he says technology genuinely helps him prepare lectures, write, and guide students.

The irony is sharp. The same technological progress that assists a blind professor is also the tool students used to game his assessment. At that intersection, what is at stake is not just one class’s grades. A university’s reputation sits on the table too.

What this means beyond Providence

This case travels well beyond Brown University. Universities in Indonesia and across the world are already navigating generative AI — from essay assignments to undergraduate theses. Many instructors are redesigning assessments, but clear institutional policies and reliable detection tools remain scarce.

In large classes, the temptation to use AI to speed through an assignment arrives quickly. Students under deadline pressure can be tempted. Instructors often only notice when the work looks too polished, too uniform. That is the core problem: technology moves fast; academic regulations tend to lag.

The lesson from the Brown AI cheating case is therefore pointed. Banning AI in a course handbook is not enough. Universities need assessments designed to resist manipulation, explicit guidance on ethical AI use, and clear consequences when boundaries are crossed.

Brown has not yet announced specific disciplinary outcomes for the students allegedly involved. But the figures already on record are hard to minimize: out of 89 midterm takers, 40 scored a perfect 100 — and 22 of the 27 students who skipped the in-person final came from exactly that group.

What comes next at Brown will signal how seriously Ivy League institutions are willing to confront AI cheating in the open, rather than handling it as a quiet internal matter. Other universities will be watching.

Key Takeaways

① Brown economics professor Roberto Serrano identified evidence that roughly 50 students used ChatGPT during a March midterm, with 40 of 89 students scoring a perfect 100.

② After Serrano announced an in-person final, the class average dropped from 96 to 48 — and 22 of the 27 students who skipped the final had previously scored perfect on the midterm.

③ In response, Serrano eliminated take-home exams and weekly-assignment grades entirely, arguing universities must redesign assessments rather than rely on prohibitions alone.

FAQ

How did Serrano detect AI cheating? He identified answer patterns that matched ChatGPT outputs when the same questions were entered into the tool, alongside a statistically improbable distribution of perfect scores.

What did Brown University do? Officials described the case as a “wake-up call” after it reached the Academic Code Committee. Formal sanctions for individual students have not been publicly announced.

Does this affect Indonesian universities? Directly, no — but the dynamics are identical. Any institution using take-home or essay-based assessment is exposed to the same risk without clear AI-use policies and redesigned evaluation methods.