world 6 min read

AI Cheating Scandal Rocks Korea's Meritocratic Exam System

A perfect LEET scorer admitted to Seoul National University faces mounting evidence he taught an AI to convert exam diagrams into text — while unable to explain DNA basics himself. What comes next for standardized testing worldwide.

  • Artificial Intelligence
  • South Korea
  • Education
  • Academic Cheating

The Perfect Score That Shouldn’t Have Been Possible

It begins with a question no exam proctor can answer: when an AI-generated response looks indistinguishable from genuine brilliance, how do you know the difference?

In South Korea, that question just became impossible to ignore. A student who scored a perfect mark on the Law Eligibility Exam — the LSAT equivalent that decides who gets into law school — also placed third nationwide on the College Scholastic Ability Test (CSAT), Korea’s brutal two-day gauntlet that determines university admission. He has since enrolled in Seoul National University’s dental medicine program. And now there is a growing body of evidence suggesting he may have used artificial intelligence to cheat his way through both exams.

The allegations are not vague. They are specific enough to have drawn a formal criminal complaint from a practicing attorney, and they point to something far more sophisticated than glancing at a phone during a test.

What the Evidence Actually Shows

On November 18, 2025, attorney Bang Min-woo of law firm Min filed a complaint with the Gwanak Police Station in Seoul, accusing the student — referred to in court documents by a placeholder surname, Lee — of obstructing official duties related to both the CSAT and the LEET.

The case file contains a detail that, taken alone, would be unsettling. In the months before legally changing his name, the person identified as Lee posted on a Korea University bulletin board seeking a private tutor for Life Science II, a notoriously difficult course in the Korean high school curriculum. His stated reason: he needed help understanding the scientific columns referenced in exam questions.

A current Korea University student named G agreed to the arrangement. According to the complaint, G provided roughly an hour of tutoring on Life Science II concepts and received payment transferred under Lee’s previous name.

Here is where the story diverges sharply from anything resembling ordinary academic struggle.

During those sessions, G reported that Lee did not simply ask about biological content. Instead, he repeatedly asked how to convert exam figures, diagrams, and charts into precise textual descriptions — asking again and again until he could articulate the method flawlessly. He had G review and correct his own written descriptions multiple times. This was not someone trying to learn biology. This was someone learning how to describe images so precisely that a machine could process them.

Even more damning, G stated that despite all this tutoring, Lee still did not understand fundamental concepts like the structure of DNA or the meaning of nucleotide base sequences. Yet when the CSAT was administered, Lee answered every question in Life Science II correctly except for one.

The gap between his demonstrated knowledge and his test performance is not merely large. It is unbridgeable.

Why This Matters Beyond Korea

South Korea already carries a reputation for extreme educational pressure. The CSAT is a national event. Schools close on exam days. Parents hold vigils outside testing centers. A top score can determine lifetime earnings and social standing. The system is built on the premise — cruel as it sometimes feels — that raw, unmediated meritocracy is the only fair way to allocate scarce opportunity in a country where a single exam outcome can reshape a family’s trajectory for generations.

An AI-assisted cheater does not just break a rule. It breaks the contract.

But the mechanism described here is worth studying globally. It is not the simple phone-under-the-desk scheme that schools have spent decades trying to police. It involves pre-exam preparation designed to teach an AI system how to interpret visual data in the exact format the test requires — then presumably accessing those AI-generated responses during the exam itself. If investigators confirm this occurred, it represents a qualitatively new form of academic fraud.

This is not a student who studied harder. This is a student who trained a machine to replace the need to study at all.

The Real-Time Demonstration That Never Happened

Facing the swelling allegations, Lee announced he would publicly solve a live Korean language section of the CSAT on the same day the exam was administered, doing so at an education institute in Gangnam, Seoul. No record exists of him following through. Either the demonstration never materialized, or it produced nothing that would dispel doubt.

That silence speaks louder than any public performance ever could.

Who Wins. Who Loses.

If the allegations prove true, several groups lose. Students who spent years grinding through after-school cram sessions without access to AI tools see their sacrifice devalued. Universities face reputational damage when admissions processes appear gamed. The legal profession, whose gatekeeping exam is at the center of this scandal, will feel particular sting given the LEET’s role in filtering entrants to one of the country’s most powerful careers.

Someone wins only if the system responds decisively. An investigation that concludes with a guilty verdict, a revoked admission, and criminal penalties sends a message that the rules still matter. An investigation that produces ambiguity — or no investigation at all — signals that whoever can best leverage AI tools has already won, regardless of fairness.

The world is watching South Korea because the stakes here are uniquely visible. Other countries administer standardized tests in quieter ways. Few generate the same national obsession. But the method described in these allegations — training an AI to transcribe and interpret visual exam materials before the test day — is replicable anywhere an AI tool exists and a proctor cannot see what a student sees.

What Comes Next

The police investigation is underway. Whether criminal charges follow depends on evidence that survives scrutiny. A civil suit or administrative review by the testing organizations could proceed independently. And even before any legal resolution, the cultural impact is already spreading.

Parents are re-evaluating what counts as legitimate preparation. Educators are confronting the possibility that AI has created a tunnel between practice and performance that conventional proctoring cannot seal. Testing organizations face pressure to redesign exams around skills AI cannot replicate — a tall order when the very tools they use to grade may themselves be vulnerable.

The most uncomfortable implication may be the simplest one: perfection is no longer proof of ability. When a student cannot explain DNA but scores near-perfect on a science section, the score has stopped being a reliable signal. It has become, instead, a record of resource access — who could afford the right kind of tutoring, who could afford the right kind of AI, who could afford to game the system before anyone noticed.

Meritocracy was always partially myth. But myths only work when people believe they are mostly true. After this scandal, belief will require proof — not just from prosecutors and investigators, but from testing organizations willing to redesign their systems from the ground up.