Emergency alerts expose cheaters: South Korea catches 111 in AI exam crackdown

After a cheating scandal involving a micro camera and a chatbot, South Korea sent nationwide disaster alerts during its most important university entrance exam. Hidden phones went off right in the exam rooms. The alerts caught 54 cheaters, and 111 were caught in total.

Industry Nov 21, 2026 7 min read Source: ITHome (CN)
Illustrative image, AI-generated

What's going on

South Korea tried something unusual at its national university entrance exam. On Thursday, November 19, 2026, authorities sent emergency disaster alerts to every phone in the country, several times over. It's the same system that normally warns of earthquakes or severe storms. Chinese tech outlet ITHome describes it as a "level 0" message. The idea was simple: if someone didn't hand in their smartphone and smuggled it into the exam room, the loud alert tone would give them away. According to the Korea Herald, 54 test-takers were caught this way. Add the other checks and the total comes to 111.

The crackdown follows a wave of AI-assisted cheating. Shortly before the exam, a scandal around the tax accountant exam had shaken the country.

The exam that brings a whole country to a halt

The College Scholastic Ability Test, or CSAT (Suneung in Korean), takes place every year on the third Thursday of November. CNBC says around 550,000 students sat it this year. The exam lasts about eight hours, from 8:40 a.m. to 5:45 p.m., and often shapes the rest of a student's life.

The whole country adjusts around it. According to CNBC, banks and the stock market opened an hour late, at 10 a.m. The stock market closed at 4:30 p.m. During the English listening section, no planes were allowed to take off or land for 35 minutes, which affected 130 flights. Government offices and companies were asked to start work at 10 a.m. or later so rush-hour traffic wouldn't hold up test-takers. Police clear the roads and drive late students to their exam sites, and military drills are suspended.

The trigger: a micro camera plus a chatbot

It started with a confession a few days before the exam. The month before, the number of top scores on the national tax accountant exam had jumped suspiciously. According to the Korea Herald, 66 candidates scored more than 90 points in Introduction to Accounting. In 2025, only two had done so, and in 2024 just one.

The top scorer, 34-year-old Han Won-seok, then admitted to cheating in a video on his lawyer's YouTube channel, wearing a mask. According to Bloomberg, he photographed the questions with a micro camera and sent the images to an AI program. He read the answers off his smartphone. In other words, he used generative AI, software that produces text and answers on its own. The Guardian quotes him as saying:

"I cheated using electronic devices and generative AI. Passing the exam with the top score was not the result of honest effort."

According to AP, Han came first in both rounds: in April he scored 397.5 out of 400 points, and in the essay round in July 663 out of 800. He had earlier denied online accusations that he wore AI glasses. He says he used the same method on another accounting exam. Police are investigating, and the finance ministry wants to revoke his license. Under Korean law, he can't retake the exam for five years. Bloomberg reports that the presidential office called the case a "serious matter" and held an emergency meeting that same Sunday. AP notes there were earlier cases this year: in June and October, test-takers wore AI glasses during an English test and university essay exams.

According to the Yonhap news agency, President Lee Jae Myung said that AI-assisted cheating in a country where state exams decide who gets opportunities shakes "the foundations of our society by undermining fairness and trust."

The package of measures

The government responded with a whole set of measures. CNBC reports that Lee wrote on X:

"Even the most brazen test-taker wouldn't be foolish enough to risk getting caught when a hidden communication device rings during the exam, would they?"

Other measures, according to CNBC and The Next Web (TNW):

Measure Details
Proctors around 75,000 proctors at 1,312 test centers, plus 8,400 staff for device checks and signal sweeps
Glasses every pair of glasses is checked; at least one specially trained proctor per site can spot AI glasses
Clothing stickers over buttons to block hidden cameras, plus watching for suspicious movements at sleeves and collars
Rewards at least $75,000 for tips on cheaters
AI companies to hand over data on users who photograph exam questions and have AI answer them; a law on this is planned
Penalties the retake ban is set to rise from one to ten years, and serious cases are to be prosecuted

TNW points out there are limits, too. Signal jammers are illegal in South Korea. Metal and radio detectors were only planned on a limited scale, and thermal cameras are considered unlikely because of privacy concerns. Proctors can only search bags with the owner's consent. There's also talk of requiring manufacturers to make AI devices blink or beep while in use. Education Minister Choi Kyo-jin rejected postponing the exam, saying it would cause "even greater disruption."

How exam day went

The first alert went off at 8:30 a.m., ten minutes before the start. It exposed 31 test-takers with banned devices. Their results were voided on the spot, even though the exam hadn't even started yet. Two more alerts at 2:21 p.m. and 2:57 p.m. caught 23 more, for 54 in total. Using metal detectors, Wi-Fi signal sweeps and manual checks, police and proctors found 52 more. Five test-takers were caught in the act of using banned devices. That adds up to 111 cases, according to Yonhap.

A death overshadowed the day. In July, Lee Jin-woo had become the first person ever to get a perfect score on the LEET law school aptitude test, and he later published a book about his study methods. After Han's confession, people online accused him of cheating with AI as well. Two lawyers filed criminal complaints for obstruction of official duties. On exam day, he was supposed to prove his abilities in a YouTube interview but never showed up. After he was reported missing, police found him dead in a motel at 1:12 p.m. The preliminary cause of death is cardiac arrest. According to Yonhap, a note at the scene said he was physically and mentally exhausted. It didn't mention the cheating allegations. None of the sources cite any evidence that he cheated.

The big picture

The case shows how cheap AI cheating has become. A micro camera and an ordinary chatbot were enough to finish first on a demanding professional exam. South Korea is responding with drastic measures, from disaster alerts to forcing AI providers to hand over data. That last one in particular raises privacy questions. It's still unclear what exactly the agreement with AI companies looks like and which companies are involved.

Other countries face the same problem. According to TNW, Norway plans to ban AI glasses in schools, and some US universities are returning to in-person exams. On November 11, England's higher education regulator warned in a report, based on a survey of more than 300 students and staff, that AI detection software is unreliable. Supervised exams may have to replace take-home coursework. Whether alert tones and glasses checks will be enough in the long run is doubtful, because the devices keep getting smaller. And the death shows how quickly a mere suspicion online can turn into a witch hunt.

Related videos

111 Students Caught Cheating with Electronic Devices During Suneung... 13 More Added Overnight / YTN YTN · YouTube
Suneung 2026 concludes, leaving AI cheating prevention and fairness as key challenges Arirang News · YouTube

Sources

#south korea#ai cheating#exams#education#smart glasses
Read the original source

Ask the AI assistant

Questions about this article, installation or background? Our assistant knows this page and runs on our own machine, not in someone else's cloud.

Connecting …

The Assistant is an AI (Gemma 4) and can make mistakes – please check important details against the source. Questions are used only to generate an answer and are deleted after one hour. Please don't enter any personal data.

Comments

What do you think? Already tried it? Write it here, no sign-up needed.

  1. Loading comments …

Comments are public. Links aren't allowed, and we delete abusive posts. You can delete your own comment at any time in this browser. Learn more in the Privacy Policy.