Mexico’s flagship university forces 58,000 applicants back to supervised pen-and-paper tests.
Mexico’s National Autonomous University of Mexico gambled on technology this year. For the first time, its high-stakes undergraduate entrance exam went fully online. Nearly 160,000 hopefuls took the multiple-choice test from home, monitored by AI-powered webcams, facial recognition, lockdown browsers, and remote human proctors. The results exploded. Scores that once looked exceptional became commonplace. Top marks surged fivefold. Suspicions of widespread cheating followed. Now the university is sending tens of thousands of students back into exam halls with paper and pens.
Score Spike That Broke Trust
Between 2021 and 2025, just 3.5 percent of applicants scored 100 or higher on the 120-question exam. This year that figure hit 16.3 percent. The share scoring 110 or more jumped from 0.9 percent to 5.5 percent. In medicine, one of the most competitive tracks, almost 30 percent cleared 100 points, triple the prior rate.
UNAM initially invalidated about 3,000 exams for clear irregularities such as identity swaps or external help. A technical commission of experts then reviewed the full data. Statistical modeling suggested far broader problems—potentially involving nearly half the test-takers through AI tools, leaked questions sold online, second monitors outside the camera frame, hidden earphones, or proxies.
Cheating tips had circulated openly on social media before the May-June testing windows. The AI systems and one human supervisor for every 150 students simply could not keep up.
The Forced Return to Analog
University rector Leonardo Lomelí accepted the commission’s recommendation for an in-person “control exam.” Roughly 58,000 applicants—those who scored high enough to claim or nearly claim a seat under historical cutoffs—must sit the new test between August 12 and 19. Venues include Mexico City, León, Oaxaca, and Tijuana so students from other states do not all travel to the capital. Only about 22,000 places exist. New-student classes were delayed until August 31.
Lomelí apologized to honest applicants who now face a second high-pressure test. “The UNAM knows how to recognize mistakes and self-correct,” he said. “There is no evasion or complacency.” The university also filed criminal complaints against individuals and companies allegedly selling exam assistance.
What This Reveals About AI in Education
UNAM’s experiment aimed at greater access. Students no longer needed to travel long distances. The tech stack looked robust on paper: lockdown browsers that blocked other apps, AI watching for phones or extra people, encrypted sessions. In practice it exposed the limits of current remote proctoring against determined users armed with generative AI and simple workarounds.
The episode lands at a moment when universities worldwide wrestle with the same tools. AI can generate answers in seconds. Detection systems lag. Multiple-choice formats prove especially vulnerable because pasted answers leave fewer forensic traces than essays. Honest students pay the price when systems fail—extra stress, delayed starts, eroded trust.
UNAM’s pivot back to supervised paper testing is temporary but telling. It prioritizes fairness over convenience for one of Latin America’s most competitive admissions processes. The larger question remains open: how institutions design assessments that stay valid when powerful AI sits in every student’s pocket. For now, Mexico’s top university has chosen the oldest technology available—ink, paper, and human eyes in the room.
AI Disclosure: This article was created with the assistance of artificial intelligence tools and was reviewed and edited by the Glowls News editorial team before publication.