AI Fraud at Brown University: The Crisis of Academic Integrity in the LLM Era

Mass AI Fraud at Brown University

Professor Roberto Serrano of Brown University has uncovered evidence of widespread AI-driven cheating in his advanced undergraduate mathematical economics course (ECON 1170), marking one of the largest academic integrity scandals in the Ivy League. At least 50 students are believed to have used artificial intelligence to cheat on a March midterm exam, a discovery that has prompted a call for a fundamental shift in how elite universities evaluate student learning.

Professor Serrano, the Harrison S. Kravis University Professor of Economics, detected the fraud after a take-home, closed-book midterm resulted in an extraordinary average score of 96/100, with 40 students achieving a perfect score. Grading assistants noted unusual passages in the answers that mirrored outputs generated by ChatGPT. The scale of the fraud became evident when Serrano transitioned the final exam to an in-person format; the average score plummeted to 48/100, and 22 of the 27 students who skipped the final had previously scored a perfect 100 on the midterm.

The Failure of Take-Home Assessments

The transition to AI-enabled cheating has rendered traditional "take-home, closed-book" exams obsolete, as these formats provide an environment where deception is both easy and highly remunerative. Professor Serrano had implemented the take-home format for the first time in 34 years of teaching, intending to reduce student anxiety following a campus shooting in December that had left two dead and nine injured.

This incident reflects a broader trend across elite institutions. Princeton University recently ended a 133-year tradition of unproctored exams—originally based on an 1893 Honor Code—returning to in-person proctoring to combat the ease of AI-assisted deception. In response to the fraud in his own course, Professor Serrano has announced two immediate changes for the upcoming academic year:

1. Elimination of take-home exams, regardless of their theoretical appropriateness.
2. Removal of weekly exercises from the final grade, as these are now easily completed using AI.

Institutional and Systemic Responses

The response from university administration to mass AI fraud is often characterized by silence or a reluctance to act, potentially influenced by the influence of wealthy donors. Professor Serrano reported the fraud to high-ranking officials at Brown, including the university president and dean, but received little to no response until the case reached the Academic Code Committee. Serrano argues that the university must publicly admit the seriousness of the situation to preserve the prestige and utility of higher education.

Perspectives on Academic Integrity and Grading

Discussion among academics and students suggests that the problem is not merely the tool (AI), but the systemic incentives surrounding grading and credentialism:

• Adversarial Course Design: Some educators are moving toward "adversarial" curriculum design. For example, at Dartmouth, some CS professors are implementing 1-on-1 interviews to verify that students actually understand the code they submit, regardless of whether it was AI-generated.
• The "Arms Race" of Cheating: Students in competitive, curve-graded programs may feel forced to use AI to remain competitive if they suspect their peers are doing the same.
• The Credentialism Critique: Some argue that the focus on grades is a "farce" of credentialism, where the prestige of the institution matters more than actual learning, reducing the incentive for students to engage with the material honestly.

"If we no longer defend truth and decency and honesty, then what kind of credibility are we going to have as academics?" — Professor Roberto Serrano

Strategies for AI-Resistant Evaluation

To maintain academic standards, educators are exploring a mix of assessment methods that prioritize process over output. Based on current academic discourse, three primary strategies have emerged:

• Punishment: Detecting AI use on homework and treating it as a standard violation of academic integrity.
• Prevention: Shifting to live, offline, or oral assessments (such as the one-on-one interview model used in some European universities) that are virtually impossible to cheat on.
• Embrace and Pivot: Assessing the process of learning rather than the final result, or designing "open-AI" exams that focus on problem definition and critical synthesis—tasks that remain difficult for LLMs to perform accurately without human guidance.