Korea's Perfect-Score Scandal and the End of Paper Exams
A LEET perfect-scorer faces police allegations of AI-assisted cheating, including over his college entrance exam record. The case exposes a crisis every nation relying on standardized tests should take seriously.
The Crack in the Foundation
Bang Min-woo did not file his complaint at Seoul’s Gwanak Police Station out of professional curiosity. The attorney represents clients who have been defrauded by public institutions and expects prosecutors to treat this case as potential obstruction of public duties. But the real weight of his filing comes from what it implies: the man at its center, Lee Jin-woo, may have used artificial intelligence to game one of the most consequential examinations in South Korea.
Lee is the self-publicized perfect scorer on the Legal Aptitude Evaluation Test, the nationwide exam that determines entry into Korea’s judicial training institute and, ultimately, careers as judges, prosecutors, and attorneys. That alone would make him a celebrity. It also makes him, apparently, a suspect.
What We Know So Far
The suspicion did not begin with the LEET. It traces back to November 2025, shortly after a tax accountant exam topping off the national ranking came forward and admitted using AI during the test. That confession opened the door to re-examining other perfect-score claims.
Last month, a post on a Korean University online community described a tutoring session. A graduate student wrote that he hired a tutor approximately a week before the 2026 college entrance examination, the CSAT, for help with Life Science II. Over roughly one hour, the tutor allegedly could not grasp fundamental concepts such as DNA structure and nucleotide sequences. The student paid 30,000 won. The remittance record showed a sender name matching Lee Jin-woo’s legal name before he changed it.
Lee has never confirmed or denied that he was the tutor. He has only addressed the name discrepancy publicly, saying he composite-edited his pre-name-change identity onto his official transcript to avoid having it disclosed externally.
If the tutoring claim holds up, it creates an uncomfortable gap between the man who supposedly needed remedial biology instruction days before a major exam and the test-taker who posted a near-perfect CSAT result showing only a single wrong answer in that same subject.
The Pattern of Improvement
Lee’s academic trajectory deserves scrutiny regardless of the AI question. His own account paints a picture of extraordinary transformation. In 2013, at age 19, he took the CSAT for the first time and received overall scores placing him in the sixth through eighth grade bands across every section. Five years later, in 2018, he still ranked in the seventh band for the Korean language section.
Then, in 2024, he passed the CSAT with top marks and enrolled at Korea University’s Department of Political Science and International Relations. In 2026, he reportedly scored first-rank performance across all subjects and gained admission to Seoul National University’s College of Dentistry.
He has attributed his turnaround to a period of analog living following a failed business. According to his public statements, he walked tens of kilometers without his phone after losing it during a nationwide journey, and credited the resulting mental clarity with restoring his cognitive ability.
Whether that narrative is genuine or carefully constructed remains unanswered. But the timing matters: it arrived in a media landscape where the idea of a prodigious score recovery sparked exactly the kind of skepticism now crystallizing into a police complaint.
Why This Matters Beyond Korea
South Korea’s entrance-exam ecosystem is unique in its intensity. The CSAT alone determines university placement for millions of students each year, and the LEET sits at the apex of a separate but equally high-stakes track leading to the legal profession. The pressure is structural. Every year, hundreds of thousands sit these exams, and a single point can shift a life trajectory permanently.
But the problem exposed by Lee’s case is not Korean. It is global.
Any country that still relies on closed-room, timed, pen-and-paper or computer-delivered standardized testing now faces the same vulnerability. Large language models can draft essays, solve problems, and generate citations in seconds. The tools exist. The question is no longer whether someone will try to use them inside an exam room—it is whether the system can detect it afterward.
The Detection Problem
Police investigation remains prospective. As of this writing, neither the LEET nor the CSAT records have been officially ruled adulterated. The complaint filed by Bang is asking investigators to assess whether applicable offenses were committed. That is a different thing from establishing guilt.
What is already clear is that existing anti-cheating infrastructure is poorly equipped for an AI-native generation. Traditional proctoring watches for eyes leaving the screen, unusual keystrokes, or unauthorized devices. It does not monitor whether answers were generated by a model and pasted in real time. Even eye-tracking software cannot distinguish between a student thinking deeply about a question and a student waiting for an LLM to output a response.
Some testing organizations have experimented with live-streamed proctoring, biometric verification, and randomized question pools. None of those measures have proven robust against AI augmentation when the examinee has access to a smartphone or wearable device—and Korea’s exam halls, despite their reputation for tight security, have not been immune to smuggling attempts in previous cycles.
What Comes Next
If investigators substantiate the allegation, the consequences would extend well beyond one examinee’s transcript. A formally verified finding of AI-assisted cheating on the LEET would force the Ministry of Justice and the Korea Judicial Research and Training Institute to reconsider how they administer and secure their selection process. A parallel finding regarding the CSAT would ripple through the Korea Educational Development Institute and potentially reshape policy for the next decade of exam delivery.
Even if the allegations fail to prove, the episode has already altered the conversation. Students, parents, and educators are now discussing openly what previously remained implicit: that perfect scores are no longer reliable evidence of mastery under conditions where generative AI is available outside the exam hall.
The harder institutional response will come eventually. Some version of open-book AI exams, oral defenses, or continuous portfolio assessment is likely to emerge in Korea and elsewhere. The question is speed. Institutions that move slowly will keep producing results that look impressive and cannot be trusted. Institutions that move too fast will face their own credibility crises.
Lee Jin-woo’s case may yet turn out to be an anomaly resolved in favor of due process. But anomalies accumulate. And by the time the pattern becomes undeniable, the damage to public confidence in the scoring system will already be done.