Nearly 160,000 applicants took the UNAM entrance exam remotely this summer using AI-powered webcam proctoring. The results showed a significant increase in high scores, raising concerns about cheating. Between 2021 and 2025, 3.5 percent of test takers scored 100 or more on the 120-question exam. This year, 16.3 percent did so, with even higher increases at the top end. The surge led to accusations of widespread cheating, prompting UNAM to form a commission to investigate. The commission recommended a control exam for all affected applicants, potentially impacting about 58,000 people.
UNAM’s rector apologized to honest applicants, acknowledging the need for the retest to ensure fairness. The exam was multiple choice, making it difficult to detect AI-based cheating. However, reports suggest students used cheat sheets, leaked questions, and other traditional methods. A New York Times report noted cheating tips circulating before the exam, including advice on hiding earphones and using AI models. The university also used LockDown Browser from Respondus and Territorium’s AI proctoring system, but these measures failed to prevent cheating. UNAM canceled nearly 2 percent of exams for unspecified issues, indicating some irregularities were detected.
UNAM acknowledged the probability that many applicants received help during the test. AI expert Raul Rojas suggested statistical modeling indicated nearly half of the students were cheating. As a result, almost 58,000 applicants will retake the exam in official settings with human proctors.
Source: arstechnica
