The AI Educator home

Press release ·

South Korea's AI exam cheating case: inspecting glasses will not show who did the thinking

South Korea's top tax accountant exam scorer admitted on 15 November to using a hidden camera and AI to cheat, three days before the national university entrance exam. Dan Fitzpatrick says policing devices will not settle who did the thinking, so assessment must change.

In response to: South Korea races to curb AI cheating ahead of college entrance exam (The Korea Herald)

Han Won-seok, 34, the top scorer on this year's tax accountant qualification exam in South Korea, said in a video posted on 15 November that he photographed exam questions with a miniature camera, sent them to an AI tool and checked the answers on his smartphone, the Korea Herald reported. He scored 397.5 out of 400 in the April first round, with 159 of 160 multiple-choice questions correct. In introductory tax law, where he scored full marks, 16,856 candidates averaged 30.18. He said he used the same method in the July round and in the first round of the certified public accountant exam in March. Police have his written confession, and the Ministry of Economy and Finance is seeking to suspend his qualification.

On 16 November, Prime Minister Han Seong-sook told students to hand in all smart devices before the College Scholastic Ability Test on Thursday 19 November and set up a government-wide emergency response team. The Education Ministry will place at least one trained official in each of the 1,312 exam centres to help proctors spot AI glasses. It said it is still working out how to detect miniature cameras, and it ruled out blanket physical searches and a postponement.

This matters beyond Korea because ranked exams decide real places and jobs. In the tax accountant exam's second round, 722 of 6,606 candidates passed, a rate of 10.92%. When a result can be produced by a hidden camera and an AI tool, every honest candidate is ranked against it. Two proctors were assigned to Han's room, according to the Korea Herald, so the problem is not that nobody was watching.

Dan Fitzpatrick's analysis is that this is the limit of surveillance as a strategy. Devices will keep getting smaller and every inspection rule is a patch for the last method. The lasting answer is to decide what each assessment is meant to prove. A closed-book exam can still prove recall under secure conditions. It cannot prove that someone can reason, explain and judge, and that is what an employer or a university is trusting when it reads a grade. The same question applies to coursework and homework in schools and colleges everywhere.

What leaders should do now: school and college leaders should list which assessments really must be secure and invigilated, and keep those. For everything else, assess the process as well as the product, with drafts, oral questioning and pupils explaining their reasoning in class. Tell pupils and parents plainly what AI use is allowed in each task. Employers that rely on exam-based qualifications should ask how a candidate shows competence beyond a single sitting.

Dan Fitzpatrick advises schools and education bodies on AI and writes on AI and the future of education. His comment on this story draws on that work with school leaders and educators.

Dan Fitzpatrick is available for interview and comment on this story. Available today for live or pre-recorded radio, TV, podcast and print interviews, with broadcast-quality remote setup for live or pre-recorded down-the-line interviews. Contact: dan@theaieducator.io

“You can inspect every pair of glasses and still not know who did the thinking. Exams built on recall under surveillance will lose that race to a camera and an AI tool. Keep secure checks where they matter, and redesign the rest so pupils show their reasoning in conversation, in drafts and out loud. Outsource the doing, not the thinking, and make sure the thinking is what we assess.”
Dan Fitzpatrick, The AI Educator

About Dan Fitzpatrick

Dan Fitzpatrick is an internationally recognised keynote speaker, five-times bestselling author and Forbes contributor, leading the charge on safe and innovative AI adoption. An educator and former senior leader, Dan advises schools, governments and organisations around the world, working with leaders across the United States, UK, Middle East and beyond.

Dan is available for interview and comment. Media contact: Dan Fitzpatrick, dan@theaieducator.io.

More statements

Expert comment

Need a comment today?

Dan responds to media requests the same day where he can.

Contact Dan