Via hobletsonthego.com
Mexico's largest university saw suspiciously high scores after its first fully remote entrance exam, and the fallout is a case study in why digital verification still has a long way to go
When nearly 160,000 students sat down to take UNAM’s entrance exam this summer, they did something no previous class had done: they took it entirely from home. AI-powered webcam proctoring, a lockdown browser, facial recognition. The whole nine yards of digital surveillance.
The results were, to put it gently, implausible.
In UNAM’s Medicine program, one of the most competitive undergraduate tracks in Latin America, 29.3% of candidates scored 100 or higher out of 120. The year before, that figure was 9%. That’s not grade inflation. That’s a statistical earthquake, and it has forced the university to summon roughly 58,000 applicants back for control exams while a commission of 16 experts investigates what went wrong.
What happened at UNAM #
UNAM, formally the National Autonomous University of Mexico, is Latin America’s largest public university with over 373,000 students. Its entrance exam, the Concurso de Selección, is legendarily competitive. Admission rates in popular fields like Medicine historically hover around 1.5%.
For 2026, UNAM shifted the entire exam to a remote format for the first time. The stated goal was accessibility: let students take the test from wherever they are, reduce logistical costs, and modernize the process. The university partnered with a company called Territorium Life, whose AI proctoring software monitored test-takers via webcam using facial and voice recognition. The exams ran over several weeks from late May through early June. Results dropped in July. And almost immediately, the numbers didn’t add up.
The Medicine scores were the most glaring red flag, but the anomalies weren’t limited to one program. UNAM initially annulled only about 2% of exams, but suspicions of AI-assisted cheating extended far beyond that small slice. The university has since d further enrollments and launched an internal review.
The AI proctoring paradox #
Here’s the thing about AI proctoring: it was supposed to prevent exactly this. Lockdown browsers restrict what students can access on their devices. Facial recognition confirms identity. Webcam monitoring watches for suspicious behavior.
In practice, the system monitored only a single device. If you had a second laptop, a phone, or a friend with a language model open in the next room, the proctoring software was essentially watching the front door while the back door was wide open.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our