Roberto Serrano, an economics professor at Brown University in Providence, Rhode Island, discovered massive cases of fraud in his course in the spring semester of 2026. As the specialist portal Inside Higher Ed reports, the majority of his class presumably used generative artificial intelligence to achieve top grades on an intermediate exam.
The incident puts the fundamental weaknesses of current examination formats at the center of the higher education policy debate. At the same time, the hesitant reaction of the university administration shows how unprepared academic institutions are to react to the systematic use of systems like ChatGPT by their students.
Out of compassion into the fraud trap
The starting point for this development was a tragic, exceptional situation at the East Coast University. After a deadly shooting on campus in December 2025 in which two people died, Serrano wanted to reach out to his traumatized students.
He converted the midterm exam of his demanding mathematical economics course, usually written in the lecture hall, into an unsupervised assignment. As the Spanish daily newspaper El País writes, an unusually large number of interested parties then registered for the course, which in retrospect can be seen as the first indication of the planned fraud.
The result of the intermediate examination was unprecedentedly high with an average grade of 96 out of 100 possible points. Historically, the average for this exam, which the lecturer made even more demanding this time due to the format, was usually only between 65 and 80 points.
When correcting, the professor and his team noticed an extremely awkward and unnatural style of argument in the answers. When they fed the tasks themselves to ChatGPT, the software spat out exactly the same convoluted solutions that the students had previously submitted in batches.
A drastic drop in grades as proof
To confirm his suspicions, Serrano did not cancel the intermediate exam directly, but simply changed the conditions for the final exam. He announced that this final test would again take place under supervision in the lecture hall and that the homework would only be included in the evaluation if the distribution of grades was approximately identical.
The class’s reaction was clear, as 18 participants dropped out of the course immediately and nine others didn’t even show up for the exam. Of these 27 absentees, 22 had previously achieved full marks in the uncontrolled intermediate examination. For the remaining participants, the in-person exam ended in academic disaster. The grade point average plummeted to 48.6 percent, which Serrano said was an all-time low for his course.
Top Article
${content}
${custom_anzeige-badge}
${custom_tr-badge}
${section}
${title}
A total of 19 people failed the course because their actual knowledge was obviously not sufficient to answer the questions without digital aids. The lecturer then declared the intermediate examination invalid and weighted the in-person exam at 80 percent for the final grade.
The institutions are capitulating to technology
However, the administrative processing of the case is difficult and does not reflect well on the faculty. The responsible committees at Brown University reacted very cautiously to the reported numbers and instead asked the professor to file a formal complaint against each individual, including digital evidence.
Serrano strictly rejected this approach because established tools for detecting artificial intelligence are known to be unreliable and often produce incorrect results. He criticizes the passivity of those responsible and calls for an honest debate about the problem. He made his fundamental criticism of this mentality clear to Inside Higher Ed in drastic terms: “We cannot afford a society in which a significant portion of our best young minds think that cheating is okay. We cannot allow ourselves to become idiots.”
A systemic problem
The case from Rhode Island is an example of a much broader development in the US educational landscape. For example, the renowned Princeton University in the state of New Jersey recently abolished its 133-year-old code of honor for unsupervised exams because the principle of trust can no longer be maintained in the age of generative text models.
Even though digital tools undoubtedly enable new methodological approaches in education, they also reveal significant risks for the traceability of individual performance. Universities obviously have to develop fundamentally new evaluation concepts if academic degrees are to continue to serve as a reliable indicator of acquired knowledge in the future.
