AI Glasses Solved Four CSAT Math Questions in Under 10 Seconds

A pair of Rokid AI glasses answered four mock CSAT math questions correctly in under 10 seconds during a hands-on trial at an optical store in Seoul's Myeong-dong shopping district. A separate calculus problem — a definite integral — took less than two seconds.

The demonstration lands days before Korea's College Scholastic Ability Test, scheduled for Nov. 19, and it shows exactly why exam authorities are nervous. Consumer AI wearables have reached the point where a student can read solutions inside their own eyewear faster than they can read the question itself.

What happened in the Seoul store?

A reporter looked through a pair of Rokid glasses at a page of four math questions from a mock version of the CSAT. The answers appeared inside the lenses before the reporter had finished making sense of the first question. All four were correct.

"This is the very product that has been in the controversy lately over cheating," the optician said.

The mechanics are simple, and that simplicity is the story. The Rokid glasses use a built-in camera to capture a question, send it to an AI system through a connected smartphone, and project the generated answer back onto the lenses. There is no fumbling with a phone under a desk. The whole loop runs through a device the wearer is already looking through.

Takeaway: the hardware doesn't just make cheating easier — it makes it nearly invisible to anyone watching the test-taker's hands.

How big is the problem already?

This is not hypothetical. Data from the Human Resources Development Service of Korea, disclosed in October during a parliamentary audit, showed five cases involving AI smart glasses at state-run qualification examinations this year. Four involved Chinese-manufactured Rokid products. One involved glasses from U.S. tech giant Meta.

Five detected cases at state-run exams in a single year means the devices have already moved from novelty to enforcement problem. And the CSAT, which determines university placement for hundreds of thousands of Korean students, is by far the highest-stakes test on the calendar.

Takeaway: the parliamentary record shows this is a current enforcement problem, not a future one.

What does the gear cost, and who can get it?

The Rokid glasses retail for 999,000 won — about $750 — in Korea. But you don't need to buy them. One rental service advertises rates as low as 9,900 won a day, roughly the price of a couple of coffees.

That rental price matters more than the sticker price. A $750 device is an impulse purchase for some families and a stretch for others. A 9,900-won daily rental puts the cheating toolkit within reach of essentially anyone, timed conveniently to exam season.

Takeaway: the low rental price turns a niche gadget into a mass-access risk for a single test day.

How does the Meta option compare?

The trial also covered a pair of Ray-Ban Meta glasses. Unlike the Rokid model, the tested pair had no display. The user speaks to it and listens to replies through built-in speakers.

Meta sells a separate display-equipped model in the United States, but it is not officially available in Korea. That leaves the audio-only version as the practical Meta option for now — and it's a worse tool for discreet cheating. Whispering a question and listening to a spoken answer in a silent exam room is conspicuous in a way that quietly reading text inside a lens is not.

That hardly rules out misuse. It just means the Rokid model, with its silent visual output, is the sharper concern for proctors today.

Takeaway: the display, not the AI, is what makes Rokid the bigger exam-room risk.

Where does the system fall short?

The Rokid glasses have real limitations, and the trial surfaced them. At one point, the system indicated it could not see the question clearly and needed a better view. At other times, it stopped responding entirely and appeared to wait for another order rather than follow earlier instructions.

Open questions remain. Would an instruction like "show only the correct answers, without explanations, for the next 30 minutes" hold for the full period? How long does the battery last under repeated use? Neither was tested. But the raw speed of the demonstration — correct answers in seconds — suggested that even a brief, partially functional window of use could carry real cheating value.

Takeaway: the device is imperfect, but it doesn't need to be perfect to be useful for a few critical minutes.

Can proctors actually catch these glasses?

The glasses may betray themselves. The Rokid frames are thicker than ordinary eyewear, and the lenses carry a noticeable, distinct reflection. In a quiet shop, with a pair in your hand and time to look, the differences stand out.

A crowded exam room is a different environment. A proctor managing dozens of students under time pressure has seconds per candidate, not minutes. A thick frame reads as a style choice; a faint lens reflection is easy to miss at a distance.

The trial proved neither that the glasses can get someone through an exam undetected, nor that they can't. What it proved is narrower and more unsettling: the answers arrive fast, they arrive silently, and they arrive inside the one object nobody expects a student to surrender — their own glasses.

Takeaway: detection is a visual-judgment problem under bad conditions, and the device's telltales are subtle ones.

What's at stake on Nov. 19?

The CSAT is Korea's single most consequential exam, and the fairness of its administration carries enormous social weight. Five detected smart-glasses incidents at state-run qualification exams this year, plus a retail ecosystem that rents the hardware by the day, give proctors a problem that existing check-in procedures were not designed for.

The Myeong-dong trial was brief and uncontrolled, and it shouldn't be read as an accuracy benchmark. But it documented something any exam authority can verify in ten seconds at a shop counter: the answers show up, they show up inside the lenses, and they show up fast enough to matter on test day. What happens on Nov. 19 will test whether Korea's exam rooms can spot them.