AI glasses in exams: when cheating beats the top 5 of the class, the assessment system has a bigger problem

By Zendoric · June 28, 2026.

The sequence is almost cinematic: a student in Taiwan aspired to enter an elite medical school, and the invigilators caught him not through a confession or a tip-off, but because his glasses gave off heat. In South Korea, two candidates for an English-proficiency exam became the country's first documented cases of cheating using AI glasses. Meanwhile China, with its Gaokao —the world's most massive university-entrance exam, with more than ten million participants a year— imposed the mandatory inspection of every candidate's frames before entering the room.

These are not isolated anecdotes: they are the symptom of a technology that has matured faster than education systems have been able to anticipate. And there is data backing it with uncomfortable precision. Assistant professor Meng Zili, of the Hong Kong University of Science and Technology (HKUST), put commercial smart glasses through a simulated university-level electrical-engineering test. The device visually captured the question, transmitted it to a large language model and projected the answers directly onto the lenses. The result: AI-assisted performance ranked among the top five in a class of more than a hundred students, clearly beating the average grade of 72. It is not a marginal exploit; it is a systematic competitive advantage.

What makes containment especially difficult is the direction of the market. Meta launched its AI Ray-Bans in 2023 and, according to the data gathered by the article, more than seven million units were sold last year. Each new generation of these devices is thinner, more discreet and capable of running more powerful models with less reliance on an external connection. Detecting a wireless earpiece was already hard; detecting glasses that look conventional and are distinguishable only by the heat they generate is another level of operational challenge.

Thomas Corbin, a professor at Deakin University in Australia, put it bluntly: «Wearable AI poses the same challenge for exams that ChatGPT posed for written assignments in 2022, and I don't think there is any reliable way to maintain current exam practices». The comparison is apt. The arrival of chatbots in 2022 forced a redesign of written tests and, even so, four years later the debate over the authenticity of academic work remains unresolved. Glasses transfer that same problem to the physical environment of the classroom.

Institutional responses, for now, are reactive in nature: inspection of frames in China, working groups in South Korea, review of protocols in Taiwan. They are reasonable short-term measures, but insufficient if the technology keeps evolving at the current pace. The underlying problem is not technical; it is conceptual.

Here the most valuable perspective comes from the very researchers who exposed the vulnerability. Professor Zhang Jun, co-director of the HKUST study, noted that the real question is not how to ban the glasses, but «how quickly we can rethink and adapt the education system, how we change the way we teach and how we assess students». Kong Siu Cheung, director of the Centre for Education in AI and Digital Competency at the Education University of Hong Kong, goes further: he proposes that AI should not be treated only as a threat but as a reality we have to learn to live with, reinforcing in students critical thinking, reasoning and metacognitive skills rather than memorization as an end in itself.

This connects with a debate the education sector has been putting off for years. If an exam mainly measures the ability to retain and reproduce information, and if an accessible and ever-cheaper tool can do that better than any student, the exam is measuring something that has ceased to be relevant. It is not an argument for eliminating assessment, but for redesigning it around what AI still cannot trivially replicate: contextual judgment, original argumentation, the ability to question the model's own output.

The East Asian case has a specific cultural dimension that amplifies the urgency. In countries where a single exam can determine access to university, the profession and, to a large extent, social status, the pressure on students is extraordinary. That pressure does not disappear by banning gadgets; it finds new outlets. Cheating with smart glasses is, in part, a symptom of that pressure poorly managed over decades.

What these incidents confirm is that the window for an orderly transition is closing. Each exam cycle that passes without structural reform is a cycle in which the gap between what the test measures and what truly matters widens. The region's education authorities have before them an opportunity —and also an obligation— to lead a change that, if it does not come from within, will end up being imposed by technology from outside.