On a crisp September morning, the Convention Center at Sejong University in Seoul was filled with the low hum of laptops, excited whispers, and the occasional burst of laughter. It was September 10, and the “2026 AI Companions Project” had brought 170 middle and high school students—arranged in 35 teams—from across Korea to a hackathon unlike any other. The event was the Responsible & Safe AI division of a broader initiative hosted by the Ministry of Science and ICT and organized by the Korea Foundation for the Advancement of Science and Creativity. Dongascience, a science-focused media company, was in charge of planning and running this particular division, and it had chosen a theme that spoke directly to the lives of teenagers: the fight against AI-generated fake news. This was not a theoretical exercise, nor a simple essay contest. Students were thrown into what the organizers called an “AI Fake News Defense Battle,” a joint mission in which they had to design and build their own AI detectors capable of telling real science news from fabricated stories. To make the challenge feel authentic, Dongascience drew on 100 carefully selected, professionally processed cutting-edge science articles from its own archives, all of which served as the raw material for the hackathon problems. The hackathon was also notable because it focused on the AI ethics issues that teenagers actually face, from deepfakes and AI hallucinations to the subtler problem of misinformation that sounds completely plausible. Two AI specialist companies provided crucial technical support: Playmore built a dedicated environment on its AI platform LOOM, where participants tackled 36 fake news detection problems, and Upstage offered free access to its large language model, Solar. With those tools in hand, students were not merely playing with models or copying code; they were wrestling with the same uncertainties and ambiguities that professional fact-checkers and AI researchers face every day. It was hard, it was messy, and it was exactly the kind of real-world learning that classrooms often struggle to provide. The fact that a media company had designed the entire experience gave it an extra layer of urgency—because journalists, after all, know better than most how quickly a false story can spread.
From the very first words of Dongascience Deputy CEO Joo Se-hoon, it was clear that this hackathon was meant to be about more than competition. He looked at the young faces in the audience and admitted that, from the perspective of a media company that had been producing articles for a long time, the fake news generated by today’s AI no longer felt like someone else’s problem. It had become an inescapable part of the information landscape. He then invited the students to explore together the very question he and his colleagues grapple with every day: “How should AI distinguish fake news?” But he did not hand them a textbook answer. Instead, he encouraged them to experience firsthand the difference between leaving everything to AI without any guidance and providing clear, thoughtful criteria that the AI can follow. He stressed that the results would be strikingly different in each case—and that thinking with their teammates about why those results came out as they did would itself be a major learning experience. Around the room, a group of “Youth Mentors”—a carefully assembled team of undergraduate and graduate students majoring in AI, along with working professionals from the field—moved from table to table. They did not simply fix bugs or hand out solutions. They asked questions, offered technical coaching, and quietly shared career advice with students who were only beginning to imagine what it might be like to work in artificial intelligence. Their presence transformed the hackathon from a contest into a mentorship experience. The learning was reinforced by two special talks. Professor Minseok Lee of the College of Software Convergence at Kookmin University delivered a keynote that deepened the students’ understanding of how AI systems learn, where they fail, and why they sometimes hallucinate confidently wrong answers. Then Playmore CEO Kim Jin-jung, whose company had provided the LOOM platform, took the stage to give a special lecture on the theme “How to Work with AI.” Drawing on years of experience in an industry that has been forced to adapt quickly to rapid advances in AI, he shared insights that felt practical, honest, and deeply relevant to teenagers trying to find their place in a shifting technological world.
By the final stretch, the hackathon had evolved into something far more meaningful than a coding exercise. During their final presentations, students stood before judges and peers to explain not only what they had built, but what their AI detectors had gotten wrong and why. This was perhaps the most striking part of the day. Many teams independently diagnosed the causes of incorrect answers: AI hallucinations, where the model confidently invents facts; misinterpretations of context; and subtle tricks of language that made false stories look credible. Instead of proudly showing off flawless performance, they demonstrated a critical eye toward AI outputs—exactly the mindset that responsible AI use demands. Their willingness to analyze their own tools’ failures showed a level of intellectual maturity that impressed everyone in the room. That critical spirit was rewarded in several categories. Twelve teams, including Team 1 and Team 3 from Geochang Daeseong High School, received the Fake News Detection Award, presented as the Chairman’s Award of the Korea Foundation for the Advancement of Science and Creativity. Another twelve teams, including the Underflow team from Sangil Media High School, earned the Presentation Award, given as the CEO’s Award of Dongascience. And eleven teams, including the AI Guardians team from Nagok Middle School, took home the Team Synergy Award, presented as the CEO’s Award of Playmore. Awards, of course, always make a day special, but they did not capture the full range of what happened in that room. The real evidence of success was visible in the students’ faces as they walked through their slide decks and demonstrated their detectors. They had entered as teenagers with a vague sense that AI could be misleading; they left as young investigators with a practical toolkit for spotting falsehoods and a new vocabulary for discussing algorithmic risk. The hackathon had been designed to make them think, and the awards simply recognized those who had thought most deeply.
The most moving testimony came from students whose relationship with technology had been genuinely transformed. One striking example was Cheongpung Middle School, a school where the entire student body participated in the project. Student Lee Geon-ho spoke candidly about his earlier experiences. He had used AI frequently in daily life, he admitted, but he had often been disappointed when it returned answers that were completely off the mark. Through the hackathon, he said, he finally learned how to write effective prompts—how to give clear, specific instructions and criteria that can turn AI from a frustrating oracle into a useful collaborator. What excited him even more was the process of building a program that could distinguish fake news on its own. He described the experience as fascinating, and it had lit a spark in him. He wanted to stay interested in artificial intelligence and keep exploring how to use it ethically in the years ahead. Another student, Kwon Ji-su, emphasized the human dimension of the event. For her, the hackathon was a meaningful time in which she became closer with friends, seniors, and juniors as they interacted and communicated around difficult problems. She was grateful for the university student mentors who stood by her team whenever they got stuck, patiently guiding them through moments of frustration and confusion. Without that support, she felt, they might never have made it to the finish line. Her last words were simple and telling: “I definitely want to take part again.” In an age when technology can feel isolating, the hackathon had managed to be deeply human. It brought young people together around shared challenges, taught them to rely on one another, and reminded them that learning—even about AI—is ultimately a social act.
Behind the students, too, were teachers who had watched the entire journey unfold. Kim Kyung-jin, a teacher at Cheongpung Middle School who supervised the participating students, offered a remarkable glimpse into the preparation. She explained that all the students at their small rural school had received division training in the “Responsible & Safe AI” project during the first semester, then decided to take on the challenge of the AI Companions hackathon together. In the weeks before the competition, the students showed great passion as they repeatedly practiced writing prompts until the skill began to feel natural. On the day of the finals, many of them were visibly nervous, especially in the early stages. But Kim described the deep sense of reward she felt watching them solve problems on their own, gain confidence with each correct answer, and ultimately experience a strong sense of achievement. For a teacher, moments like that are rare and precious. She ended with a heartfelt thank-you to everyone involved in creating what she called a wonderful arena where her students could grow into future talents. The event’s impact, however, was not meant to end when the hackathon closed. Dongascience announced plans for participants to carry out independent inquiry activities together with the youth mentors after the competition. Applying the principles they had learned during the hackathon, each team would run its own small project to solve problems in its own neighborhoods and schools. This was where the real-world relevance of AI literacy would be tested. Instead of a one-day event, the hackathon was becoming the beginning of an ongoing process of exploration. These projects would culminate in a demo day scheduled for November 22 at Osco in Cheongju, where students would present their findings and share what they had discovered. The entire arc—from training to competition to community investigation—showed a thoughtful, holistic approach to science education that honored both technical skill and ethical awareness.
Looking back at the “2026 AI Companions Project” hackathon, what stands out is how thoughtfully it addressed the anxieties and promises of the current technological moment. We are living in a time when AI-generated images, voice clones, and convincingly fake articles are no longer science fiction; they are part of everyday digital life. Teenagers, who often spend more time online than anyone, are especially vulnerable to being fooled—and especially likely to encounter the ethical gray zones of AI. Yet this event refused to treat them as passive victims. Instead, it treated young people as capable problem-solvers who could learn to build defensive tools, ask probing questions, and hold AI accountable. The partnership between Dongascience’s journalistic resources, Playmore’s platform, Upstage’s language model, and the dedicated mentors created a rich learning ecosystem. Students did not simply read about AI ethics; they experienced the messy, frustrating, joyful process of turning ethical principles into working software. They learned to be skeptical without being cynical, to trust evidence over vibes, and to collaborate under pressure. By the end of the day, they had not only created detectors for fake news; they had strengthened their own internal detectors as well. That is a victory that extends far beyond a single hackathon. It suggests that the next generation will not merely inherit the challenges of the AI era but will be equipped to shape them. The students from Cheongpung Middle School and the dozens of other teams across Korea returned home with more than certificates. They returned with curiosity, confidence, and a sense of responsibility. And with the planned follow-up projects and the November demo day on the horizon, their learning journey is only just beginning. In the struggle between truth and misinformation, these young defenders of fact are proving that the human mind, properly guided, is still the most powerful algorithm of all.