Brown University AI cheating scandal exposes gap between allegations and enforcement

Brown University economics professor Roberto Serrano alleges that at least 50 students used AI to cheat on his midterm exam, sparking what he calls the largest known AI cheating scandal at an Ivy League school. The case reveals a growing gap between detecting AI-enabled cheating and actually enforcing academic integrity policies.

What happened

After a deadly campus shooting in December, Serrano moved his in-class midterm and final exam to an online, take-home format for the spring semester. Course enrollment jumped to 86 students, far exceeding the previous record of 30. The online midterm averaged 96, compared to the typical range of 65 to 80. Forty students achieved a perfect score, and Serrano detected answers that resembled ChatGPT outputs.

Serrano then announced the final exam would return to an in-person format and that midterm scores would be voided if they did not match the final average. Only 59 students took the final exam, which averaged 48.6 with a maximum of 95 and a minimum of zero. Of the 27 students who dropped the course and final, 22 had received a perfect midterm score.

Serrano published an op-ed in The Free Press on June 28 detailing his findings. He said he provided evidence to Brown's Standing Committee on the Academic Code but received no acknowledgment until the story went viral. Brown University stated it has been "consistently responsive" and that the committee is now moving forward with formal adjudication after Serrano provided the necessary details on July 8.

Why AI builders should care

This incident is not just an academic story. It illustrates a structural problem that will affect any organization deploying AI tools in assessment, credentialing, or certification contexts. When AI can produce plausible answers for any text-based exam, the entire model of unsupervised assessment breaks down.

For builders creating AI detection tools, proctoring software, or assessment platforms, the Brown case highlights a clear market need: solutions that help institutions move from detection to enforcement. The gap between identifying cheating and actually adjudicating it is where most organizations get stuck.

Practical implications

Educational institutions will need to redesign assessments to be less susceptible to AI-assisted manipulation. In-person components, oral exams, and project-based evaluations are likely to become more common. Formal adjudication pathways, like Brown's Standing Committee on the Academic Code, will be invoked more often as AI-related allegations grow.

For product teams building AI integrity tools, the key insight is that detection alone is not enough. Institutions need workflows that connect detection data to formal adjudication processes, with clear evidence packages and audit trails that can withstand scrutiny.

Caveats

Details are still developing. The case is under formal investigation by Brown's Standing Committee on the Academic Code, and no final ruling has been published. The numbers reported come from Serrano's own analysis and media reports, not from an independent audit. Different outlets may report varying interpretations of the same data. The university's timeline of events differs from Serrano's account, and the full picture will depend on the committee's findings.