Nearly 58,000 applicants to Mexico's National Autonomous University (UNAM) — Latin America's largest university — must retake the entrance exam after its first fully remote, AI-supervised sitting produced results so suspicious that the institution invalidated them.

Around 160,000 applicants took the test under a "lockdown browser" and AI-powered webcam proctoring designed to flag cheating by analyzing gaze, background activity and browser behavior. Instead, the system appears to have missed widespread cheating: the number of top scores rose roughly fivefold compared with previous years, according to Ars Technica.

UNAM said irregularities and an atypical score distribution undermined confidence in the results. The roughly 58,000 students who must retake the exam include applicants who had already secured places — a painful consequence for students who thought their admission was settled.

The debacle is a high-profile example of the risks of automated proctoring, which universities embraced during remote learning because manually monitoring thousands of students is expensive. Critics have long argued such systems cannot reliably distinguish cheating from normal movement, poor lighting or connectivity issues — and that even small error rates become devastating at scale. UNAM says the retake will be held in person.

Sources