Exam Cheating with AI Smart Glasses and Micro-Devices: 2026 Detection Guide

AI smart glasses can capture exam questions through built-in cameras, process them with AI, and display answers directly on the lenses — all while looking like ordinary eyewear. In 2026, confirmed cheating incidents have spread across South Korea, Taiwan, China, and the UK, prompting the first-ever institutional bans from the College Board, Seoul’s education office, and national exam systems. This guide explains how the technology works, which models are in use, what schools are doing to respond, and the practical detection methods available right now.

Key Takeaways

AI smart glasses are real and actively being used to cheat — students in South Korea were caught wearing them during TOEIC exams, and one in Taiwan had their medical school entrance exam nullified after proctors noticed their glasses frame was hot from processing.
Major institutions have already banned them — The College Board prohibited all smart glasses from SAT testing starting March 2026. Seoul’s education office ordered a city-wide ban on smart glasses during final exams in June 2026.
The technology is mainstream, not niche — Over 7 million pairs of Meta-Ray-Ban smart glasses sold in 2025 alone, making these devices mass-market consumer products rather than obscure gadgets.
Detection requires a layered approach — No single method catches everything. Effective detection combines physical screening, behavioral analysis, Bluetooth scanning, and proctor training.
The accessibility dilemma is real — Students who need smart glasses for vision correction create enforcement challenges that institutions are still working through.

What Are AI Smart Glasses and How Do They Work for Cheating?

AI smart glasses are eyewear with built-in cameras, microphones, processors, and displays. Unlike traditional smart glasses that simply project notifications or provide audio feedback, the AI-enabled models can capture exam questions visually, process them through large language models, and project answers onto the inner lenses — all hands-free and nearly invisible to proctors.

The workflow is straightforward:

Capture: A student wearing smart glasses looks at an exam question. The built-in camera captures the text or handwritten problem.
Process: The image is sent to a connected AI model — either running locally on the device or routed through a cloud connection. Large language models generate an answer in seconds.
Display: The answer appears on the inner lens as a heads-up display, or is transmitted through bone-conduction audio to the student’s ear. The student reads or hears the answer without touching a keyboard or looking at a screen.

A student with this technology looks exactly like a student taking a normal exam. There’s no keyboard, no screen, no obvious device interaction. The glasses resemble ordinary eyewear.

Which Models Are Available in 2026?

Several AI-enabled smart glasses models are commercially available and have been documented in cheating incidents:

Rokid AI Glasses: Tested in the HKUST controlled experiment and used in South Korea’s TOEIC cheating cases. Available on Chinese secondhand platforms for as little as 40–80 yuan ($6–$12) per day through rental services.
Xiaomi AI Glasses: Another Chinese model frequently rented through secondhand platforms for exam cheating.
Even Realities G1: A minimalist HUD-style smart glasses designed for everyday wearable use. Even Realities also published one of the most comprehensive visual detection guides for identifying smart glasses.
Meta-Ray-Ban: Meta launched the category with its first AI-enabled glasses in late 2023. The brand has since released several versions and sold over 7 million pairs in 2025 alone, making smart glasses mainstream consumer products rather than niche gadgets.

A businessman in Shenzhen told Reuters he rented Rokid and Quark AI glasses to more than 1,000 people over four months. The barrier to access is far lower than most educators assume.

Real 2026 Incidents That Changed Policy

The shift from theoretical threat to enforced reality happened through a series of documented cases across multiple countries.

South Korea: TOEIC Cheating and Criminal Prosecution

In May 2026, two test-takers were caught wearing AI-enabled smart glasses during separate TOEIC (Test of English for International Communication) sessions. Both examinees had their scores invalidated and were suspended from taking the test for four years. These were South Korea’s first reported cases of cheating with AI glasses.

The response escalated quickly. In June 2026, South Korea filed its first criminal prosecution over AI smart glasses cheating — marking the first time exam cheating became a criminal offense in the country’s history. Schools in Gyeonggi province borrowed metal detectors from local education offices. Middle schools issued emergency notices to parents stating that smart glasses are banned from exam rooms and will be treated as cheating.

Taiwan: Medical School Disqualification

A prospective student taking a National Taiwan University (NTU) medical school entrance exam was discovered wearing AI smart glasses after proctors noticed the student staring oddly at the test paper at an unusually close range. Inspection revealed the glasses frame was emitting significant heat, indicating active processing. NTU gave the applicant a null score and reported the incident to Taiwan’s College Entrance Examination Center, which has since revised its testing protocols.

China’s Gaokao: Maximum Enforcement

China took the most aggressive physical screening of any standardized test in history. For the gaokao — the country’s annual college entrance exam taken by more than 10 million students each year — authorities mandated screening of all glasses at exam sites. Smart glasses and all electronic devices were already officially banned, but the scale of enforcement was unprecedented.

UK: Ofqual Warning

Sir Ian Bauckham, head of England’s exam regulator Ofqual, warned in June 2026 that “the vast majority” of students “wouldn’t dream of cheating,” but the small minority who do “have always set about trying to subvert the system.” He specifically mentioned “biros that have apparently invisible mini video screens built into them” as a concern and said the regulator had to “move really fast because technology is moving fast.”

The HKUST Proof-of-Concept Experiment

The strongest published evidence that AI glasses can work for exam cheating comes from the Hong Kong University of Science and Technology (HKUST). Assistant Professor Meng Zili noticed a pair of stylish glasses worn by a student during an exam he was proctoring. As a researcher working on AI glasses, the frames caught his attention.

The glasses turned out to be ordinary, but the incident inspired him to partner with electrical engineering professor Zhang Jun to test commercial AI glasses in an undergraduate electrical engineering exam. The result was striking.

By simply looking at the exam paper, the glasses transmitted questions to a connected large language model, which generated answers and displayed them on the lenses. The score generated with the device placed it in the top five of over 100 students, far exceeding the class average of 72. The research team described AI glasses as a “viable technology” for handling exams.

“After conducting the experiment, it really raises the question of how much knowledge students actually need to memorize for exams, versus whether we should allow them to use AI during assessment,” Meng said.

Zhang Jun noted that with how fast technology is evolving, education is finding it hard to keep up — “every teacher feels that.”

Institutional Bans and Policy Responses

The 2026 incidents prompted some of the fastest institutional policy responses seen in recent years.

College Board Ban (March 2026)

The College Board banned all smart glasses from SAT, PSAT, and AP testing starting in March 2026. The policy is explicit in the prohibited devices list:

Any eyewear featuring integrated cameras, microphones, video displays, or internet connectivity is prohibited.
Even prescription smart glasses are banned. Students who rely on them for daily vision must bring a standard, non-connected pair of backup glasses.
Turning the device off or disconnecting it does not make it permissible. The physical presence of cameras, recording lenses, or radios violates the protocol regardless of power state.
Proctors collect smart glasses alongside mobile phones and digital watches prior to the start of testing.

This was one of the earliest and most decisive institutional responses to AI glasses cheating.

Seoul Education Office Ban (June 2026)

In June 2026, Seoul’s education office ordered a city-wide ban on smart glasses during final exams — the first government-mandated ban at the school district level in Korea. The order covered all schools’ final exams.

Courts and Military

Philadelphia’s First Judicial District barred all smart and AI-integrated eyewear from its buildings effective March 2026, with criminal penalties. Military installations across the United States have also implemented restrictions.

The Accessibility Dilemma

These bans create a genuine conflict. Students who need smart glasses for vision correction (such as patients with low vision who use augmented reality glasses for visual assistance) face a dilemma: bring backup non-connected glasses or potentially miss the test. The College Board’s approach — allowing backup non-connected pairs while banning connected ones — sets a precedent that other institutions are adopting.

Any detection policy must balance cheating prevention with the needs of students who genuinely require assistive technology. Intensive scrutiny of eyewear, hearing aids, and medical devices may discriminate against students with disabilities, and schools need policies that address smart device threats without creating inequitable surveillance.

How Schools Can Detect AI Smart Glasses and Micro-Devices

Detection is a layered problem. No single method catches everything. Here’s what each method catches and how to implement it.

1. The 60-Second Visual Checklist

The most practical immediate tool is a visual inspection checklist based on the framework published by Even Realities. A quick physical check covers four areas:

Physical Hardware Cues:

Thicker temples: Side arms look wider or heavier than normal glasses to hide batteries and processors inside.
Tiny apertures: Look near the hinge or frame corners for small microphone or camera pinholes.
LED indicators: Check the front corners for indicator lights that activate during use or recording.
Charging contacts: Look for exposed magnetic dots or metal pins on the inner temple tips.

Behavioral Signs:

Temple taps: Watch for the user repeatedly swiping or tapping the side of their frame.
Glance patterns: Notice sudden pauses where the user stares fixedly into space to read a micro-display prompt.
Voice prompts: Listen for low-voiced interactions or quick commands given to an assistant.

The complete guide, with annotated visuals showing where to look, is available at the Even Realities smart glasses detection page.

2. Ear-Scanning Technology for Micro-Device Detection

Hidden earpieces present an equally urgent threat. As tiny as 3mm — smaller than a bean — they sit deep within the ear canal and use electromagnetic induction loops that bypass standard scanner thresholds.

Duolingo’s patent-pending ear-scan technology, announced at DETcon London in June 2026, addresses this gap. The process:

The Duolingo English Test mobile app collects sensor data while the phone is held close to each ear.
A visual distance scan points the phone camera at each ear so it is centered on the screen.
A close-range check moves the phone close to the ear (mimicking a phone call) while an algorithm processes sensor and proximity data.
The sequence is repeated for each ear to ensure no unauthorized in-ear listening devices are present.

For K-12 and in-person exams, schools can implement similar procedures by using the mobile app as a quick verification step before exams begin.

3. Bluetooth Scanning

Bluetooth scanners act as radio-frequency sentinels. During an exam session, they periodically activate passive observation of the local Bluetooth spectrum and analyze Received Signal Strength Indicator (RSSI) metrics. Weak signals are categorized as background noise (devices in neighboring rooms). Strong signals exceeding a specific threshold are identified as being within arm’s reach of the student.

In a 1.3 million-test implementation for the National Driving License program, the system achieved a 100% prevention rate of impersonation attempts involving hidden earpieces — the closest thing to a technical solution for invisible hardware.

4. Proctor Training

Knowing a device exists doesn’t mean invigilators can recognize it. The BBC reported that UK invigilators are now being specifically trained to spot covert equipment — not just as policy, but as an operational requirement. Training should cover:

Visual inspection techniques (using the 60-second checklist above)
Behavioral awareness (unusual gaze patterns, micro-pauses, temple tapping)
Physical screening procedures (metal detectors, backpack inspections)
When to escalate (what to do when a device is suspected)

What Should Educators Do Now?

Here’s a practical decision framework for responding to the AI glasses and micro-device threat.

Immediate Actions (This Week)

Update your device ban. Smart glasses, micro-earpieces, and Bluetooth trackers are already in active use. Your exam policy needs to name them explicitly and ban them. Follow the College Board’s lead: ban connected eyewear but allow backup non-connected glasses.
Train your invigilators. A written ban doesn’t help if nobody can recognize the device. UK invigilators are now being trained specifically to spot covert equipment — this should be a requirement at your institution. Use the Even Realities detection checklist as the training material.
Add physical screening. South Korean schools borrowed metal detectors from local education offices. China required full eyewear screening for 10 million students. Physical screening is the simplest but most universally applicable tool.

Medium-Term Actions (This Semester)

Add behavioral monitoring. If you’re using proctoring tools that only monitor the student’s computer, you’re blind to hardware-based cheating. EduLegit’s Video Guard platform captures and analyzes behavioral signals in real time — micro-movements, sustained reading behavior, gaze coordinates, and keystroke dynamics. This catches the human tells left by hardware cheating.
Evaluate your AI detector strategy. Relying on AI text detectors alone for cheating detection is a fundamental category error. A study of 14 AI detection tools found false-positive rates as high as 50%. If you’re using AI text detectors without behavioral analysis and version history, add those layers. Pair any AI detection with the EduLegit AI Content Detector and behavioral analysis.

Long-Term Planning

Redesign assessments. Thomas Corbin at Deakin University told CNN that wearable AI presents “as much of a challenge to exams as ChatGPT did to essays in 2022.” He doesn’t believe there is any reliable way to maintain current exam practices going forward. The long-term response may not be better metal detectors or smarter proctors. It may be a redesign of assessment itself — oral examinations, project-based evaluations, or open-book formats that test application rather than recall.

Why This Matters: The Bigger Picture

The cheating gadgets flooding exams in 2026 represent a structural shift, not a trend. The devices are cheaper, smaller, and harder to detect than anything that preceded them. The data — from HKUST’s controlled test results to the criminal prosecution in South Korea — proves that AI-assisted cheating is already widespread at institutions that thought they were insulated.

Thomas Corbin’s warning — “If we’re seeing a few cases being reported, we’re seeing a lot more cases not being reported” — suggests actual incidence is significantly higher than public reports. The testing bodies are moving. The data is clear. And the students you’re assessing are competing in an environment where the devices exist — whether your institution detects them or not.

But despite being a major disruption to education, the power of AI should not be a reason to discourage its use. Professor Kong Siu Cheung at the Education University of Hong Kong noted that educational systems should focus on developing students’ thinking capabilities and metacognition to prevent over-reliance on AI. “We should use technology. We should use AI. We should not just say avoid it, stop using it… The bottom line is: don’t outsource your thinking capability.”

Related Academic Integrity Guides

Cheating Gadgets and Technologies in 2026: How to Detect Them — A broader guide covering all cheating gadgets including earpieces, nano-cams, smart pens, and hidden screens
Online Exam Security: 10 Proven Methods — Defense-in-depth strategies for preventing cheating in online exams
Video Guard — Behavioral analysis platform that captures human tells left by hardware cheating
AI Content Detector — AI-generated text detection with version history verification

What You Can Do Next

If you need to discuss how these detection layers fit into your institution’s existing setup, the EduLegit contact page is the right place to start. The team of educators can help you evaluate tools and strategies that work for your specific context.

If your priority is monitoring for honest exams in real time, EduLegit’s Video Guard platform captures behavioral signals, keystroke dynamics, and gaze patterns that hardware cheating leaves behind. It doesn’t just record — it analyzes.

Recommendation: Implement a layered detection approach (physical screening + proctor training + behavioral monitoring) immediately. Don’t rely on AI text detectors alone for hardware-based cheating.

Key tradeoff: Security versus accessibility. Intensive screening of eyewear, hearing aids, and medical devices must be balanced against students who genuinely need assistive technology. The College Board’s approach — ban connected glasses but allow backup non-connected pairs — sets the right precedent.

Common mistake: Assuming that turning off a smart glass device makes it permissible. College Board’s policy is explicit: the physical presence of cameras, Bluetooth radios, or recording lenses violates the rule regardless of whether the device is turned off.