Mexico's largest university is forcing roughly 58,000 applicants to retake its entrance exam in person after a remote, AI-proctored version of the test produced a suspicious surge in high scores and widespread allegations of cheating. The decision by the National Autonomous University of Mexico, known as UNAM, is one of the most dramatic fallout cases yet from the rapid adoption of AI-powered exam surveillance in higher education.
The case highlights the growing tension between institutions rushing to adopt automated proctoring and the reality that AI monitoring systems can be circumvented. For more on how AI is reshaping institutions, see our breaking AI news.
What went wrong
Earlier this summer, nearly 160,000 applicants took UNAM's entrance exam entirely remotely — a first for the institution. Over several weeks from late May through early June 2026, students completed the 120-question test from home using a "lockdown" browser and AI-powered webcam proctoring software.
The results bore little resemblance to past years. Between 2021 and 2025, 3.5 percent of test takers scored 100 or more on the exam. This year, 16.3 percent did so. The spike was even more pronounced at the very top: between 2021 and 2025, just 0.9 percent of applicants scored 110 or higher, but this year 5.5 percent reached that mark. In effect, top scores roughly quintupled.
That statistical anomaly triggered accusations of widespread cheating and prompted UNAM to convene a commission of experts to investigate. The panel has now submitted its recommendation.
The control exam
The commission's conclusion is blunt: the only way to restore confidence is to require applicants to sit for a new test in person. UNAM will administer what it calls a "control exam," and it will apply not only to those who secured a spot based on this year's results, but also to everyone who would have been admitted based on minimum successful scores since 2021. Approximately 58,000 people could be affected.
According to Gaceta UNAM, the university's official publication, the rector apologized to honest applicants who will now have to prepare for and take the test again despite having done nothing wrong. The rector nonetheless defended the move, calling the control exam "necessary to give certainty and guarantee equity in access."
The timeline is tight. Classes are currently scheduled to begin on August 10, 2026, meaning the university will have to organize and grade a mass in-person exam for tens of thousands of applicants in a matter of days, or else delay the start of the academic term.
How the proctoring failed
The remote exam relied on two layers of security. According to reporting by El País, the test required the LockDown Browser from Respondus, which locks down the testing environment by preventing printing, copying, web browsing, and access to other applications during the exam.
UNAM also deployed a proctoring system from Territorium that used AI algorithms to monitor each test-taker's webcam. The system was designed to detect when another person had taken the applicant's place, when cell phones or earphones were in use, or when the applicant left the camera frame. Each exam also had one human supervisor for every 150 applicants, who received automated alerts and could flag irregularities.
These measures did not work well enough. UNAM ultimately canceled nearly 2 percent of total exams for unspecified conduct issues, but the vast majority of suspected cheating appears to have slipped through.
How students allegedly cheated
A report in The New York Times last week detailed the methods students reportedly used to defeat the monitoring. According to the Times, cheating tips were circulating widely before the tests went live. Online accounts advised students to position monitors outside the webcam's frame where they could access ChatGPT or other AI models. Others suggested hiding earphones under their hair, or hiring someone else to take the exam off camera.
Because the exam was multiple choice rather than essay-based, it was difficult to detect the telltale signs of AI assistance, such as complete answers being pasted into text boxes all at once. Students may also have relied on cheat sheets, leaked questions, and other traditional methods of academic dishonesty.
A cautionary tale for AI proctoring
The UNAM episode is likely to reverberate far beyond Mexico. Universities and certification bodies worldwide have invested heavily in AI-powered remote proctoring, marketed as a scalable way to maintain exam integrity during online and hybrid testing. The technology promises to replace human invigilators with always-on algorithmic surveillance, but the UNAM case suggests that determined test-takers can still find gaps.
The incident also raises uncomfortable questions about fairness. Honest applicants who followed the rules now face the burden of retaking the exam, while the university must absorb the logistical cost and reputational damage of a botched testing rollout. UNAM has acknowledged what it called the probability that the remote format was compromised, and the institution is now paying the price for prioritizing convenience over the proven reliability of in-person assessment.
For institutions weighing similar remote testing programs, the lesson from Mexico City is sobering: AI proctoring may be necessary, but as a standalone guarantee of integrity, it is far from sufficient.
Stay Ahead of AI
The UNAM case is a defining moment in the debate over AI in education. Read more AI industry coverage and Read more AI news →
