The National Autonomous University of Mexico (UNAM) is facing a crisis after its attempt to fully digitize undergraduate admissions using AI-driven proctoring backfired. Despite deploying a system designed to record every applicant's behavior and support human monitors, the university discovered an alarming number of perfect scores that pointed toward systemic cheating.

Statistical anomalies and AI fraud

The investigation revealed irregularities suggesting that students either used generative AI tools to find answers or obtained the exam questions in advance. Out of approximately 150,000 tests, around 3,000 have been invalidated. However, the fallout extends far beyond those specific cases; thousands of students are now in limbo as UNAM suspended enrollments to allow a technical commission to review the entire selection process.

Reverting to in-person testing

To restore academic integrity, especially for highly competitive degrees like Medicine—where the passing threshold among the 120 questions is highest—UNAM has announced mandatory in-person exams for selected candidates. This decision highlights a significant gap between the promise of AI surveillance and its actual effectiveness in preventing real-time malpractice during remote testing.

The AI surveillance paradox

The UNAM incident illustrates a growing technological arms race in education. While the university used AI to safeguard the process, students leveraged similar technologies to bypass those very safeguards. This failure suggests that relying solely on automated monitoring may be insufficient, pushing institutions to reconsider the fundamental structure of high-stakes assessments in an era of ubiquitous AI.