UNAM's AI Exam Surveillance Failure: A Cautionary Tale 2026

The National Autonomous University of Mexico (UNAM) recently discovered a harsh reality about AI in online exams: there is no foolproof way to prevent cheating. For the first time, UNAM held its admission exam online and deployed an AI-based surveillance system to monitor students via camera and microphone. The result was a major academic scandal, as thousands of students used AI and other methods to pass the very exam designed to stop them.

What Happened at UNAM?

UNAM, Mexico's most prestigious public university, administered its admission exam online between May and June. Nearly 158,000 applicants took the 120-question multiple-choice test, competing for just over 20,000 spots. The exam was conducted using LockDown Browser, which blocks other applications and internet access during the test. To further prevent cheating, UNAM contracted an AI surveillance system to analyze students' behavior in real time through their cameras and microphones.

When results were published on July 17, the numbers were suspicious. Between 2021 and 2025, an average of 3.5% of applicants scored 100 points or higher. This year, that figure jumped to 16.3%—nearly five times higher. Even more striking, the percentage of students scoring above 110 points rose from 0.9% to 5.5%. These anomalies suggested widespread cheating, despite the AI surveillance.

How Did Students Bypass AI Surveillance?

Despite the LockDown Browser and AI monitoring, students found ways to cheat. While the exact methods remain unclear, it's likely they exploited gaps in the surveillance system, such as using secondary devices, virtual machines, or even having another person take the exam. Some may have used AI tools on separate devices to look up answers, while others might have found ways to disable or deceive the camera and microphone monitoring.

The incident highlights a critical flaw: AI surveillance can be outsmarted, especially when it relies on behavioral analysis that can be mimicked or avoided. It also raises questions about privacy and the effectiveness of remote proctoring in high-stakes exams.

Why This Matters for AI in Education

UNAM's experience is a cautionary tale for institutions worldwide. As universities increasingly turn to AI for exam proctoring, they must recognize its limitations. AI surveillance is not a silver bullet; it can create a false sense of security while infringing on student privacy. Moreover, the arms race between cheating methods and detection technologies is likely to continue, with students finding new ways to exploit weaknesses. The longer history behind the story is traced in UNAM's AI Exam Surveillance Failure: A Cautionary Tale.

This case also underscores the need for alternative assessment methods that focus on critical thinking and application rather than rote memorization, which is easier to cheat on. It also calls for transparency and dialogue about the ethical use of AI in education. The background to this story is set out in Celtic's Epic Collapse: A Cautionary Tale in European Football. A related thread runs through First Cross-Channel Swimmer: The Cautionary Tale of Matthew Webb. Earlier coverage in Trump's Iran War Repeats Afghanistan Mistakes: A Cautionary Tale shows how the argument developed.

Key Takeaways

• UNAM's AI surveillance system failed to prevent cheating in its online admission exam.
• Score anomalies showed a fivefold increase in high scores, indicating widespread cheating.
• Students likely bypassed LockDown Browser and AI monitoring using secondary devices or other workarounds.
• The incident highlights the limitations of AI proctoring and the need for better assessment methods.
• Institutions must balance security with privacy and consider alternative exam formats.

Comparing Exam Security Measures

What to Watch Next

UNAM has not yet announced how it will handle the cheating scandal, but it may invalidate results or re-administer the exam. The incident is likely to spark debates about the role of AI in education and the need for robust, fair assessment methods. Other universities considering similar surveillance systems should take note: AI is not a substitute for thoughtful exam design and human oversight. A closer look at the same subject appears in AI Workplace Surveillance: How Employers Monitor Employees. Anyone following the wider picture can start with Ai Video: latest articles. The same theme surfaces in Student Spends Four Hours Making AI Presentation Look Human, which takes a different angle.

As AI continues to evolve, so will the methods to circumvent it. Institutions must stay vigilant and prioritize integrity without compromising student rights. Context on how this played out before sits in Mexico University Resit Exam After Cheating Scandal. A companion piece, Mexico University Resit Exam: 58,000 Students Face Cheating, approaches the same subject from another angle.