AI-assisted cheater tops South Korea’s tax accountant exam
Hundreds of thousands sat South Korea’s CSAT as authorities weighed new AI-cheating measures. A tax accountant exam cheat used a camera, smartphone and AI to rank first.

Why it matters
- Hundreds of thousands of high school students sat the CSAT on Thursday.
- A national tax accountant examinee used a miniature camera, smartphone and AI to come first.
- Measures to prevent AI cheating in the Test of Proficiency in Korean were under consideration.
- Breen says credentials act as employer proxies when hundreds of applicants make talent and judgment hard to distinguish.
- Proposed assessment formats include oral exams, projects, real-world exercises and AI use that candidates must challenge.
By Rebecca Stone
A test taker used a miniature camera, a smartphone and AI to finish first in South Korea’s national tax accountant examination, Michael Breen reports.
Michael Breen, author of “The New Koreans,” argues in an essay that presents the views as his own. He says the case exposes more than a surveillance problem. It reflects a culture that rewards narrow test performance over curiosity, judgment, creativity, character and the ability to solve problems outside school.
Hundreds of thousands of South Korean high school students sat the College Scholastic Ability Test on Thursday. They were told to leave their phones outside, keep their eyes on their own papers and follow rules intended to preserve equal conditions. For many, Breen calls the exam the most consequential few hours of their young lives.
He says authorities have been scrambling to prevent students from using AI to cheat. New measures for the Test of Proficiency in Korean were under consideration. Those steps address enforcement, while Breen argues the deeper issue lies in the culture surrounding exams.
Breen places AI within a familiar pattern. Cheating has surfaced in university admissions, civil service examinations, academic research, professional qualifications and resumes submitted for jobs or public office. The technology changes the method; it does not create the underlying incentive to gain an advantage.
What does the cheating scandal reveal?
For Breen, the problem begins with how students are trained and judged. They spend years memorizing facts, formulas, vocabulary and prescribed interpretations of texts. The examination then measures how much of that material they can reproduce within a few hours on one day, even though much of it will soon be forgotten.
The consequences of that score extend well beyond the exam hall. A student’s CSAT performance may determine which university the student attends, which companies consider an application and, ultimately, what opportunities become available later in life.
That design sends a message about what institutions value. Breen says it rewards the ability to perform under particular conditions rather than curiosity, judgment, creativity, character or real-world problem-solving ability. Students learn that one measurable result can carry more weight than a broader assessment of who they are.
Breen’s conclusion is that indifference can become reciprocal: “The exam becomes an obstacle to overcome, not an exercise whose integrity is worth protecting.” A student who believes the system does not care about the person behind the score has less reason to preserve the system’s rules.
He does not treat this as a defense of cheating. Breen calls cheating a moral failure, but one occurring inside a system with its own moral shortcomings. The failure to acknowledge those shortcomings, he argues, only makes enforcement harder.
The tax accountant case shows the endpoint of that incentive. Breen calls it cheating done so well that the test taker came top after using a camera, smartphone and AI to obtain answers.
Can stronger punishment solve the cheating?
One policy answer is to punish cheating more severely. Breen accepts that the rules must be enforced and that people who cheat should face meaningful consequences. He does not dispute the moral case for penalties or the need for exam security.
His objection is that punishment alone cannot repair a system rooted in the way society defines success. In his view, enforcement can address a broken rule without fixing the incentive that made breaking it worthwhile.
That distinction also applies to employers. Breen does not argue that standards should disappear. He argues that standards should measure demonstrated abilities rather than stand in for abilities that a university name or certificate cannot reveal.
Why do employers keep relying on credentials?
Employers face a practical information problem, according to Breen. When they must assess hundreds of applicants, they struggle to distinguish talent, intelligence, judgment and potential. A university name, grade or certificate offers an easier signal, so it becomes a proxy for quality even when it reveals little about its holder.
He recalls hearing an exasperated academic settle a disagreement with a colleague from a provincial university: “But Dr. Lee, I went to Seoul National University.” The point was not that the argument had become better. The speaker assumed the credential should settle the dispute.
Breen compares that move to a captain pulling rank on a sergeant. The implication, he says, was not that the speaker had made the better argument, but that the credential should settle it.
Some employers are already moving toward skills-based recruitment and practical assessments. Breen says these approaches allow them to hire people whose abilities can be demonstrated instead of inferred from the university they attended. The shift would let employers assess demonstrated skill rather than treat the CSAT and similar credentials as decisive proxies.
How should examinations change in the AI era?
Breen starts from a simple premise: if machines can increasingly memorize and reproduce information, an exam built around those abilities measures less of what people should be able to do.
The competencies Breen lists are judging whether an answer is reliable, separating fact from fiction, weighing competing arguments, communicating clearly and applying knowledge to an unfamiliar problem. Together, they shift the test from memory toward judgment.
His proposed design gives students access to AI during the exam. They would demonstrate how they use it, challenge its answers, identify errors and develop their own conclusions. The aim is not automated answer production; it is visible reasoning that can be assessed.
Oral examinations, project-based assessments and real-world problem-solving exercises could provide more evidence of ability than a timed test of recall. Breen presents them as alternatives that examine what candidates can do, not how quickly they reproduce a prescribed answer.
The hard constraint is scale. Any replacement must be fair, scalable and resistant to manipulation, Breen writes. A new format that cannot be administered consistently or that creates a new route to cheating would reproduce the problem rather than solve it.
The immediate policy test concerns the measures under consideration for the Test of Proficiency in Korean. Breen’s broader prescription reaches beyond preventing AI-assisted answers: employers and exam designers should judge demonstrated skill, including the ability to use and question AI, rather than rely on credentials as shortcuts.
If South Korea strengthens enforcement while leaving assessment incentives intact, Breen’s central warning remains: authorities can catch more cheaters without changing what the system rewards. His alternative demands assessments that reveal judgment and practical ability while remaining fair, scalable and resistant to manipulation.
Source: feed.koreatimes.co.kr
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
754 articles
Related articles
- AI Glasses Solved Four CSAT Math Questions in Ten Seconds
- Korea adds 8,400 staff for AI-cheating safeguards at CSAT
- South Korea tightens Suneung security after AI cheating confession
- South Korea Cracks Down on AI Cheating Before Thursday’s CSAT
- Korea to Cut Wi-Fi at College Entrance Exam Sites Over AI Cheating