Alleged mass AI cheating at Brown

Brown University economics professor Roberto Serrano has alleged an episode of alleged mass AI cheating after he converted a scheduled in-person midterm into an online, take‑home exam. Serrano told reporters the midterm produced an average score of 96 with 40 perfect scores and that many answers “resembled” output from ChatGPT and other large language models.

The claim prompted campus and national coverage and a formal review by Brown’s Standing Committee on the Academic Code. The university says it has Serrano’s materials and is “moving forward according to its procedures.”

What Serrano says happened: alleged mass AI cheating

Serrano wrote that he shifted the midterm and, later, the final to take‑home formats after a December campus shooting disrupted normal instruction. Course enrollment surged to about 86 students this year, up from roughly 30 in prior offerings, he said.

In Serrano’s account, the online midterm produced an unusually high average — about 96 — compared with typical averages of roughly 65 to 80 in previous years. Serrano reported 40 students earned perfect scores, a scale and concentration of top marks he found implausible.

He has publicly said many midterm answers bore strong similarities to ChatGPT outputs. In an op‑ed and interviews, Serrano published example answers and argued the pattern of phrasing and solution approach did not match prior student work.

“Academia is supposed to be one of our great beacons of truth,” Serrano wrote. “We cannot afford to tolerate or reinforce such appallingly low moral values among many of our best young minds.”

Midterm to final: the raw discrepancies

The course’s in‑person final, Serrano reported, was taken by 59 students and produced an average of 48.6, with a top score of 95 and a minimum of zero. That steep drop from the midterm average is central to his concerns.

Serrano also reported that 27 students dropped the course after the midterm; 22 of those who withdrew had recorded perfect midterm scores. He told the committee he would void midterm results that could not be reconciled with final performance unless students offered satisfactory explanations.

Brown’s response and investigation timeline

Brown University told reporters it was “consistently responsive” to Serrano’s concerns and said multiple academic leaders contacted him in May 2026 to explain formal adjudication options. Per the university, Serrano provided the Standing Committee on the Academic Code with detailed materials on July 8 and the committee is “moving forward according to its procedures.”

Serrano has said he first spoke with El País on June 28; subsequent coverage by Inside Higher Ed and Fox News helped bring broader attention that he said led to more formal campus action. Brown’s statement emphasizes that the committee will follow established rules for review, student notification and any potential hearings.

Investigation timeline (reported)

– May 2026: Academic leaders contacted Serrano to outline formal procedures.
– June 28, 2026: Serrano discussed the case with El País.
– July 8, 2026: Serrano provided materials to the Standing Committee on the Academic Code, per Brown statements.

Evidence, limits and open questions

Serrano’s central claim rests on patterns he says mirror LLM-generated text — phrasing, structure and the way problems were solved. He has shared examples publicly and described recurring stylistic signatures he associates with ChatGPT‑style output.

Public reporting to date, however, does not include an independent forensic confirmation of AI use on the midterm. The allegation remains subject to the university’s formal process. Outside factors could also affect observed discrepancies, including selective withdrawal, different test-taking conditions and variation in who sat for the final.

“The committee will determine whether the evidence supports academic sanctions or other conclusions,” Brown said.

Why this matters for academic integrity

Beyond a single course, the episode highlights how rapidly improving AI tools complicate assessment and honor systems. If instructors cannot reliably distinguish student work from AI‑assisted output, institutions must weigh detection, deterrence and design changes.

Policy options under discussion on campuses nationwide include revised assessment formats, clearer AI‑use policies, increased proctoring for high‑stakes tests and greater emphasis on in‑person evaluations where feasible. The standing committee’s handling of this case could influence how Brown and peer institutions respond to similar incidents.

What comes next

The Standing Committee on the Academic Code will review Serrano’s materials, allow students to respond and follow adjudication procedures if the committee finds probable cause of violations. Outcomes may range from no action if evidence is insufficient to disciplinary measures if the committee substantiates violations.

The case could also prompt broader policy changes at Brown about take‑home assessments, AI disclosure requirements and course design to mitigate future risks.

FAQ

Was cheating confirmed or is this an allegation?
This remains an allegation by Professor Serrano under formal review. Public reporting includes Serrano’s findings but no independent forensic confirmation has been published.

How did Brown respond and what is the timeline?
Brown says academic leaders contacted Serrano in May 2026. Serrano gave the committee detailed materials on July 8, and the Standing Committee on the Academic Code says it is following normal procedures.

What evidence connects exam answers to ChatGPT?
Serrano has said many answers resembled ChatGPT output and has shared sample answers and patterns he found concerning. The committee’s review will evaluate the strength of that connection.