Mail

In the spring of 2025, South Korea made a bold bet: AI-powered digital textbooks, rolled out nationwide to 3rd and 4th grade elementary and 1st grade middle school students, covering math, English, and computer science. Twelve publishers developed these textbooks, the government invested approximately 533 billion Korean won (around $850 million USD as reported), and international media called it "a pioneering experiment in AI education."

By August 2025, the Ministry of Education had stripped the AI textbooks of their official status, reclassifying them as optional supplementary materials. The whole experiment lasted four months.

What Went Wrong

Teachers in the field told a damning story. The AI failed to recognize students' handwritten numbers. It marked correct answers as wrong. It generated nonsensical responses. On top of that, serious concerns arose about how student data was being collected and protected.

Teacher Kim Cha-myung summed it up in a media interview: "Everything was rushed."

That single sentence captures the core of the failure. The AI textbooks were rolled out without adequate pilot testing. Political timelines and budget pressure outran technical validation. After a change in government, the cancellation came even faster.

The National Assembly passed legislation narrowing the legal definition of "textbook" to explicitly exclude AI-driven software.

Why This Failure Matters Globally

South Korea's case is not just a national policy stumble. At this very moment, dozens of countries are drawing up plans to integrate AI into classrooms.

The lesson experts consistently draw from this episode is straightforward:

AI should first prove itself in low-stakes settings — homework help, practice tools — before being placed at the center of core curriculum.

Putting unproven technology at the heart of learning means turning children's academic year into a live experiment. In South Korea's case, the cost of that experiment was too high.

The Right Order for Technology Adoption

This episode is a reminder of principles that EdTech advocates have long known but rarely followed in practice:

Pilot first: Test with small groups, collect data, iterate before scaling.
Train teachers before deploying tools: Technology only works if the humans using it are prepared.
Expand gradually: Start with verified functions, add complexity as confidence grows.
Protect student privacy rigorously: AI tools handling children's data require stricter standards than most commercial applications.

Hundreds of billions of won were spent on this experiment. What it leaves behind is not just failure — it is a clear, expensive, and globally useful lesson.

Sources

Rest of World, "South Korea's AI textbooks fail after rushed rollout" (2025)
WebProNews, "South Korea Axes $850M AI Textbook Program After 4 Months of Issues" (2025)
Korea Herald, "South Korea pulls plug on AI textbooks" (2025)
Futurism, "South Korea's Experiment in AI Textbooks Ends in Disaster" (2025)
South Korea Ministry of Education, AI Digital Textbooks official announcement