An AI-supervised remote exam went so badly that 58,000 students must retake it

UNAM, the sprawling public university that anchors Mexican higher education, decided this year to run its entrance exam entirely online. No classrooms, no proctors walking the aisles — just a lockdown browser and AI webcam software watching nearly 160,000 applicants sit the test over several weeks in late May and early June. The idea was convenience and scale. What it produced was chaos.

The numbers tell the story better than any accusation could. From 2021 through 2025, roughly 3.5 percent of test takers scored 100 or above out of 120 questions. This year that figure jumped to 16.3 percent. At the very top, scores of 110 or higher used to be almost vanishingly rare, under 1 percent of applicants. This year, 5.5 percent hit that mark. That's not a small bump from a slightly easier test. That's the kind of jump that makes statisticians and admissions officers immediately suspicious.

UNAM assembled a commission with a name longer than most exam sections to figure out what went wrong, and its conclusion, delivered just now, is blunt: too many people probably cheated, and there's no clean way to sort the honest scores from the inflated ones. So the university is throwing out trust in the results altogether. A new in-person