South Korea is in the middle of an AI cheating scandal that has now turned into a tragedy. On November 19, 2026, Yonhap and The Korea Times reported that Lee Jin-woo, a 32-year-old Seoul National University dental student who achieved the first perfect score on the Legal Education Eligibility Test (LEET), was found dead. He had faced suspicion that he used AI to cheat. No cheating on the LEET has been proven.
This post sticks to what mainstream outlets confirm, separates suspicion from fact, and explains what the scandal means for how AI is changing exams. It is a sensitive story involving a death, so it does not speculate about causes.
TL;DR: what is confirmed and what is not
| Question | Answer |
|---|---|
| Who was found dead? | Lee Jin-woo, 32, SNU dental student and LEET perfect scorer, per Yonhap and The Korea Times. |
| When and where? | Nov 19, 2026, 1:12 p.m., a motel in eastern Seoul, in cardiac arrest with no chance of resuscitation. |
| Was cheating proven? | No. It was alleged online, and police were investigating complaints. |
| Is there a note? | Police found a note saying he was mentally and physically exhausted. It did not mention the allegations or the investigation, per Yonhap. |
| What triggered the suspicion? | The tax accountant exam top scorer admitted AI-assisted cheating days earlier. |
| What happens to the case? | Police will continue to investigate, and the case is expected to close without prosecution referral, per Yonhap. |
If you or someone you know is struggling, contact local emergency services or a crisis line. In South Korea the national suicide prevention hotline is 109. In the United States, call or text 988.
The sequence of events
The scandal has two linked parts.
1. The tax accountant exam confession. The Associated Press reported on November 16 that Han Won-seok, 34, the top scorer on the national certified tax accountant exam, admitted in a video that he used a concealed camera to capture the questions, fed the images to a generative AI tool and read the answers on a smartphone. He had scored 397.5 out of 400 on the first-stage test in April and 663 out of 800 on the second-stage exam in July. One Korean newspaper compared that to running 100 meters in under five seconds. He said he would cooperate with police.
2. The LEET suspicion. The confession drew fresh attention to Lee Jin-woo's rise. According to The Korea Times, Lee had performed poorly on earlier CSAT attempts, later missed one question on the CSAT last year and attributed his improvement to a three-week walking trip across Korea without a smartphone. He published a book, "Brain Accel", on the methods he claimed. Two lawyers filed complaints that led the Seoul Metropolitan Police Agency to open an inquiry for obstruction of business, per Yonhap.
Lee denied cheating and said he would solve the Korean-language section of this year's CSAT in a livestream on Thursday. He did not appear, organizers cancelled the event and said his phone had been off since Wednesday. Police traced his movements through security camera footage. He was found at 1:12 p.m.
What police and officials said
Police said they will continue to investigate to establish the facts. Yonhap says the case is expected to close without referral for prosecution following his death. The Seoul Metropolitan Office of Education told Younghoon High School, where Lee took the 2026 CSAT, to preserve his question papers and answer sheets at police request, and Prime Minister Han Seong-sook directed the Ministry of Education to preserve all of last year's CSAT papers until further notice, per The Korea Times. Yonhap adds that the government is expected to consider a fact-finding team on alleged AI cheating in national exams.
None of that is a finding of guilt. Preserving papers is a way to enable review, not a verdict.
How South Korea changed its national exam
The CSAT, called Suneung, was taken on Thursday by more than 500,000 students, per the AP. The AP's pre-exam report describes the response:
- Postponement refused. Education Minister Choi Kyo-jin rejected calls to delay the exam, saying a delay would cause "even a larger disruption" to students.
- Device checks. The minister said he was confident proctors could spot AI-powered smart glasses, which typically have thick, horn-rimmed frames. Officials said each of the 1,300 test sites would get at least one specially trained supervisor, and proctors received guidance on watching for suspicious movements involving sleeves, collars or buttons.
- Options under discussion. As of the AP's November 16 report, officials were considering requiring test-takers to hand over smart devices and shutting down public Wi-Fi at test sites. The Vice Education Minister said restricting telecommunications was also being discussed, though it was unclear whether that is feasible given laws against signal jamming. Check the Ministry of Education's announcement for what was finally ordered.
- Limits. The AP also quotes an official saying metal and radio detectors would be limited, thermal cameras were highly unlikely because of privacy concerns, and supervisors cannot search belongings without consent.
The AP also lists earlier incidents: test-takers caught using AI-powered smart glasses on an English standardized test in June and on some university essay exams in October.
Why this is an AI story
The tax exam method is simple and uses ordinary tools: a hidden camera, a model that can read a photo of a question, and a phone to return the answer. Multiple-choice and short-answer exams are especially exposed because current multimodal models handle them well. That is the core of the problem for any high-stakes test taken in a room, as opposed to one taken in an environment that the proctor fully controls.
Three responses are possible, and Korea is trying the first:
- Block the channel. Cut signals, ban devices, inspect clothing. It works for a single sitting and costs a lot.
- Change the test. Move toward oral exams, in-person writing, problems that need original reasoning and verification after the fact.
- Detect after the fact. Use statistical analysis of scores and answer patterns. This risks false accusations, which is why the current case is so sensitive.
The same questions apply to hiring, certification and coding interviews. Our guide to AI curriculum for college students and for high school students discuss how schools are rethinking what they teach and test, and the Estonia AI Leap schools program shows a country building the assessment around reasoning instead of recall.
The risk of treating suspicion as fact
This case is a warning about accusation in the age of AI. A perfect score is statistically rare but not proof. The reports do not show a mechanism, a device or an AI log for the LEET. What they show is public suspicion, a livestream challenge and a police inquiry.
Responsible coverage, and responsible use of AI-detection claims, means saying what is known: someone achieved an unusual result, a confession elsewhere raised doubts, and an inquiry began. It does not mean declaring guilt. Students, teachers and employers using AI detectors should hold the same line. Detectors give probabilities, not proof, and a wrong accusation has a human cost.
What this means for builders and organizations
- Edtech and proctoring. Expect demand for verification that works against multimodal assistants, including environment control and identity checks. Be careful about marketing detection accuracy.
- Credentialing bodies. If your exam can be answered from a photo, assume a model can answer it. Redesign or add a live component.
- Employers. Take-home tests and remote coding screens face the same exposure. Our coverage of college students pushing back on campus AI mandates and the Gallup study on Gen Z and AI shows how students themselves view AI.
- Product teams. If you build vision features, think about misuse: a model that reads a photo of an exam page is a cheating tool in the wrong setting.
Questions people ask about this story
Is it true AI helped him get a perfect LEET score? That has not been shown. The reports describe suspicion that grew after a different confession, and Lee denied cheating. Police were examining complaints and said they would continue to establish facts.
Why was he suspected? His rise from poor early CSAT results to a perfect LEET score, his account of a "cognitive shift" after a smartphone-free walk, and the timing of the tax accountant confession fed public doubt, per The Korea Times and Yonhap.
How did the tax accountant exam cheater get caught? According to the AP, his unusually high marks drew suspicion after he discussed them in media interviews. He then confessed in a video on his lawyer's channel.
Can AI really answer exam questions from a photo? Current multimodal models can read an image of a question and produce an answer. The AP-reported method relied on exactly that, with a hidden camera and a phone. Accuracy depends on the subject and question type.
Could this happen outside Korea? Yes. The same tools exist everywhere. Any exam with a human proctor and no signal control is exposed to the same method, which is why other testing bodies will be watching how Korea responds.
What to watch next
- Whether the government forms a fact-finding team on AI cheating across national exams.
- Whether police publish any technical findings about the tax accountant case, which has a confession and a described method.
- Whether Korea changes exam format or rules for 2027.
- Whether other exam bodies report similar cases.
Bottom line
The confirmed facts are narrow and serious: an AI-assisted cheating confession on one national exam, a police inquiry into suspicion about another, and the death of the man at the center of that suspicion. The wider lesson is about proof. AI has made some kinds of cheating easy and some kinds of accusation tempting, and institutions need controls that work without presuming guilt.
Details reflect reports by Yonhap, The Korea Times and the Associated Press as of November 19, 2026. The situation may change.
