AI Cheating Scandal Rocks Mexico's Largest University, Up to 75,000 Exam-Takers Suspected

MEXICO CITY — The National Autonomous University of Mexico (UNAM), Latin America's most prestigious public university, is grappling with a cheating crisis after moving its highly competitive entrance exam online for the first time ever this year — complete with an AI-powered proctoring system meant to prevent exactly this kind of fraud.

More than 158,000 people sat the exam, competing for around 22,000 open spots at UNAM's main campus after its 14 affiliated preparatory schools account for the rest of each year's roughly 50,000 admissions. When results came back, the share of applicants scoring 110 or higher on the 120-point test jumped from a historical average of 0.9% to 5.5% — more than a fivefold increase that immediately drew suspicion from applicants and outside experts alike. Mathematician and AI specialist Raúl Rojas, who teaches at the Free University of Berlin and the University of Nevada, Reno, analyzed the score distribution and concluded that roughly half of all test-takers — around 75,000 people — likely cheated in some way, calling the shift in scores "not typical" of a normal exam.

Rojas said the 17-day window the online exam remained open gave ample opportunity for answers to leak and circulate, and he believes old-fashioned cheat-sheet sharing, not AI tools directly, was the primary method — though the AI-powered proctoring software, built by Mexican company Territorium Life, has drawn the brunt of public blame regardless. UNAM has so far annulled only about 3,000 of the roughly 150,000 exam results, leaving tens of thousands of students — including those not suspected of any wrongdoing — in limbo just days before the new semester was set to begin on August 10. The scandal has reached the highest levels of Mexican politics, with President Claudia Sheinbaum, herself a UNAM physics graduate, suggesting the university involve the attorney general's office to investigate.

The episode has become a global cautionary tale for institutions moving high-stakes exams online: similar AI-enabled cheating cases have surfaced in England and South Korea using smart glasses, while Australian universities were forced to abandon AI-detection tools entirely after they falsely flagged students at alarmingly high rates — suggesting the technology meant to catch AI misuse is, in practice, proving just as unreliable as the misuse itself.