In June 2026, global media outlets reported the first widespread instances of students using AI-powered smart glasses to cheat on exams across East Asia. Two examinees were caught during the TOEIC (Test of English for International Communication) in South Korea as early as May 10 and May 31, after proctors noticed suspicious eyewear before the test began. These devices allow for the real-time scanning of exam papers, utilizing built-in cameras to recognize text and instantly project answers onto the lenses, completely upending traditional methods of proctoring.
Technically, these systems integrate miniature cameras and computer vision with cloud-based access to powerful language models. A camera captures the exam page while OCR algorithms extract the text in fractions of a second, allowing the neural network to do more than just find answers—it analyzes the context, explains the underlying logic, and adapts its explanations to the specific wording of the question. This represents a fundamental shift from primitive methods like hidden earpieces or paper cheat sheets, which require significant preparation and coordination. Assistance is delivered instantaneously and no longer depends on the student having previously memorized the material.
According to various expert estimates, these detected cases are merely the tip of the iceberg. Lecturer Thomas Corbin from Deakin University in Australia, who researches the use of AI glasses in education, notes that for every registered case, there are likely many more going unnoticed. This pattern has already been demonstrated in China, where a thriving rental market for smart glasses was uncovered in early 2026, with devices being leased for $6 to $12 a day to hundreds of students and entrepreneurs specifically for exam use.
Detection is becoming increasingly difficult. While examiners rely on visual inspections and metal detectors, modern smart glasses are virtually indistinguishable from standard frames and often contain no metal. In Taiwan, a medical student was only discovered by chance when a professor noticed a strange gaze and used a thermal scan to reveal the device's heat signature—a level of scrutiny rarely seen in large-scale examinations. The current lack of standardized tests to verify that proctoring protocols are resistant to AI assistance leaves existing systems inherently vulnerable.
The high stakes in the region create a powerful incentive to utilize these new tools. In East Asia, where a single exam can dictate a student's future career and social standing, the competition is ruthless. With over 10 million applicants sitting for China’s annual national college entrance exam, authorities took the unprecedented step in June 2026 of requiring universal eyewear inspections before entry, acknowledging the scale of the problem.
Historical comparisons highlight a qualitative shift in cheating methods. Previously, technologies like hidden cameras required accomplices, complex logistics, and months of preparation. Today, a single individual needs only one device and cloud access, as the system is almost entirely autonomous. Meanwhile, countermeasures are being developed, ranging from gaze-tracking software to total bans on any eyewear in examination halls. However, their effectiveness has yet to be proven in widespread use, and the technology is evolving much faster than the systems designed to detect it.
The scale of smart glasses proliferation has already reached a critical level. Meta launched its first AI glasses with Ray-Ban in late 2023 and has since released several new versions, with more than 7 million pairs sold last year alone. Simultaneously, Chinese manufacturers such as Xiaomi, Alibaba, and Li Auto are integrating language models into their own devices, supported by government subsidies. The market shows no signs of slowing down; on the contrary, devices are becoming thinner, more discreet, and more functional.
Scientific experiments confirm the potential danger. At the Hong Kong University of Science and Technology, researcher Meng Jili tested commercial AI glasses connected to GPT during an engineering exam. The student wearing the glasses scored 92.5 out of 100—a top-five result in a class of more than 100 students, significantly outperforming the average score of 72. These results were achieved under the genuine stress of a final exam session, rather than in a controlled laboratory environment.
The problem is now extending beyond the borders of East Asia. In the United States, the College Board banned smart glasses from the SAT starting in March 2026. In Britain, the primary exam regulator Ofqual warned in June that AI glasses and micro-earpieces represent a growing threat. As early as 2024, Japan uncovered an organized cheating scandal involving TOEIC proxies where examinees photographed questions to send them to contacts via social media for answers, resulting in hundreds of invalidated scores.
This evolution is forcing educational systems to fundamentally rethink how they assess students. The focus is shifting from memorization toward critical thinking skills, the ability to defend answers orally, and the application of knowledge in novel situations—actions that are much harder to replicate using AI. At the same time, however, the issue of fairness is becoming more acute as students from affluent families gain access to these tools earlier and at lower costs, creating a new form of inequality.
Countermeasure methodologies are still in their infancy. It remains unclear how quickly detection systems can integrate eye-tracking analysis, thermal emission monitoring, or the tracking of electromagnetic signals. The coming months and years will likely see the introduction of hybrid protocols that combine traditional supervision with AI-driven monitoring and biometrics. The question is whether these will be effective at scale or if they will simply complicate life for honest students while innovative cheaters stay one step ahead.
Ultimately, educational systems are left with no choice: they must either completely overhaul the format of examinations or accept that in an era of ubiquitous AI, the concept of the closed-book exam as a validation tool has reached its limit. The first path requires courage and investment, while the second demands a reevaluation of the very meaning of certification in a society where machines already outperform humans at standard tasks.
