UNAM AI Exam Fail Forces 58,000 Students to Retake Tests in Person
When UNAM handed over anti-cheating duties to algorithms, expecting bulletproof security, human ingenuity responded with off-camera screens and hidden earpieces.
At the start of summer, nearly 160,000 applicants took the remote entrance exam for UNAM, Mexico's largest university. To keep everyone honest, applicants were forced to install Respondus LockDown Browser alongside Territorium, an AI proctoring tool meant to track webcams, flag phone usage, and monitor missing faces with a human supervisor assigned per 150 students.
The high-tech surveillance completely collapsed when the exam results came back wildly inflated. Between 2021 and 2025, exactly 3.5% of test-takers scored 100 or more points out of 120. This year, that number skyrocketed to 16.3%, while candidates scoring above 110 points jumped from 0.9% to 5.5%.
Social media guides had instructed students to position secondary monitors just outside webcam angles to query ChatGPT, conceal micro-earphones under long hair, or simply hire proxy test-takers to sit out of frame. AI researcher Raúl Rojas performed statistical modeling on the dataset, concluding that nearly half of all online candidates cheated.
In response, UNAM leadership canceled the online scores and mandated an in-person retake for 58,000 prospective students, including honest applicants who hit minimum score thresholds. Meanwhile, the cheating wave is expanding globally beyond laptops, forcing authorities in China to inspect smart glasses for over 10 million entrance test takers, while officials in the UK and Taiwan scramble to update proctoring standards for AI-enabled eyewear.
High-tech proctoring software successfully converted entrance exams into a test of stealth and peripheral camera management rather than actual knowledge. Educational institutions will keep purchasing digital panopticons until they realize that an algorithm looking through a webcam is no match for a taped sticky note on a bezel.
Source: Ars Technica
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.