A school leader sees three headlines: Estonia is bringing AI into learning, France has a framework for classroom use, and South Korea is developing AI teaching materials. It is tempting to arrange them along a single scale, from cautious to ambitious, and ask which country others should follow.
That comparison starts too early. The initiatives address different questions.
A responsible comparison of AI in schools separates access to tools, permission to use them, and education about AI itself. It then asks who the initiative serves, what has actually happened and which claims the evidence supports. Estonia, France and South Korea offer useful examples of those distinctions. They do not provide a controlled test of three competing national models.
For education leaders, the opportunity is to borrow a well-understood idea rather than a country's reputation.
First, identify what is being changed
“AI in education” can describe several interventions that may coexist.
A service can help teachers prepare resources. A pupil-facing application can support a learning activity. A national framework can define appropriate use. A curriculum can teach learners how AI works and how to question it. Professional development can prepare adults to make those decisions.
These changes have different purposes and require different evidence. A successful teacher workshop does not establish that an application improves student understanding. A required literacy course does not mean pupils have permission to use any chatbot. An announced curriculum does not show that the materials have reached classrooms.
Before comparing countries, complete a more precise sentence: this initiative changes a particular activity for a particular group, under particular conditions.
That discipline makes the three examples more informative.
Estonia: examine the support surrounding access
The official e-Estonia profile describes AI Leap as a learning-centred programme for teachers and upper-secondary students. It places teacher preparation, subject guidance and school-based professional learning communities alongside access to AI tools.
The profile also describes ITI, a tutor-style application intended to help students plan and reason through questions and explanations rather than immediately supply a finished answer. It reports expansion across upper-secondary education in 2026–27, while possible extension to vocational and basic education is being assessed.
These are programme descriptions. The profile alone does not establish improved learning or demonstrate that every interaction follows the intended design. Its full first-year report was not reviewed for this article.
The useful question for another school system is therefore broader than whether to obtain the same tool. What preparation would teachers need to understand its role? How would they recognise an unhelpful response? Where could they discuss a difficult subject-specific example?
Borrowing access while omitting that professional work would produce a different intervention. Equally, including training would not guarantee success. Leaders would still need evidence about the particular use they propose.
France: distinguish learning requirements from permission
France's Ministry of Education published its Cadre d'usage de l'IA en éducation on 13 June 2025. The ministry's web summary differentiates stages: primary pupils learn basic ideas about AI without directly manipulating generative services; limited, explained and teacher-supported classroom use begins from quatrième; lycée students may use generative AI more autonomously within an explicitly teacher-defined learning framework.
Those distinctions should not be compressed into “France allows AI from a certain age.” The educational setting and conditions remain part of the rule.
A separate service note, dated 23 January and published in the Bulletin officiel on 5 February 2026, makes dedicated Pix AI learning pathways compulsory for specified cohorts in public and contracted private establishments: quatrième, seconde and first-year CAP. Implementation is progressive across 2025–26 and 2026–27.
A requirement to learn about AI is not a universal instruction to open personal chatbot accounts. Nor does the requirement itself demonstrate that every establishment has completed implementation.
For international readers, the lesson is to separate curriculum expectations from tool-use permissions. A school can develop understanding while placing limits on direct use. Conversely, permission to use a tool does not establish that learners understand its limitations.
South Korea: ask what “AI textbooks” means
A Korean Ministry of Education communication dated 18 September 2026 describes teachers authoring AI subject textbooks and related educational resources. It distinguishes core principles in textbooks from current technological developments in supplementary materials.
Its timetable places capacity-building and development in 2026, piloting and review in 2027, and formal school application in March 2028. The latter stages are plans, not completed events as of this article's research date.
This particular announcement concerns resources for teaching AI. It should not be treated as evidence about the present operation or legal status of the separate AI-powered digital-textbook initiative.
The distinction matters beyond Korea. Teaching materials can explain machine learning without adapting themselves to a pupil. An adaptive service can use AI without teaching pupils how it works. Both may appear under a headline about “AI textbooks,” while changing very different parts of education.
When a proposal crosses languages, ask someone to explain the activity behind the label. What will teachers teach? What will pupils do? Which functions are actually automated? A familiar English phrase may conceal an important difference.
Three lessons, without a league table
The cases illuminate three questions: what professional support accompanies access, how learning requirements relate to permission, and whether AI is the subject being taught or the technology delivering instruction.
None of those questions belongs exclusively to one country. Estonia also has curriculum and governance questions; France also needs teacher support; Korea also faces implementation decisions. Selecting one initiative from each country does not describe its entire education system.
The comparison is useful because it prevents a common error: treating unlike activities as rival answers to the same problem.
It also prevents an unwarranted verdict. We cannot conclude from these sources that one country has found the best model, that another is falling behind, or that national test performance reflects the effects of these initiatives. Such claims would require different research.
Try borrowing the idea before borrowing the label
Imagine a fictional international school network planning its next academic year. Its board wants “an ambitious AI programme” and asks staff to learn from the three countries.
The first draft combines a pupil-facing assistant, compulsory literacy lessons and new teaching materials. Each component sounds reasonable. Together, however, they leave the original educational need unclear.
A more useful discussion starts with a problem: pupils accept fluent explanations without asking what supports them.
The network could respond through explicit teaching about claims and evidence, teacher-prepared examples, subject knowledge and opportunities to explain a judgment. Direct interaction with an AI system might serve part of that work, but it would require a separate justification and appropriate approval.
Staff could then examine the international examples for a particular contribution. Would collaborative teacher preparation help them design the lessons? Do they need a clearer distinction between learning about AI and using it? Would resources separating stable concepts from changing product features be useful?
These are proposals for the fictional network, not practices attributed to the three countries. The purpose is to make borrowing selective and explainable.
Follow the evidence through the change
For any chosen component, ask what would count as progress.
If the objective is teacher preparation, evidence might show whether educators can explain a proposed use, identify its limitations and adapt it to their subject. Attendance records would establish participation, but would not settle those questions.
If the objective is student understanding, the network needs evidence of what pupils can distinguish, justify or apply. A completed pathway or attractive presentation may be relevant, but it cannot carry every claim about capability.
If the objective is appropriate use, staff need to understand the conditions and recognise when to seek further advice. A policy document's existence does not demonstrate consistent decisions.
The evidence should also help leaders change their minds. If pupils can repeat terminology but cannot recognise an unsupported claim, the teaching needs examination. If teachers understand a demonstration but cannot use the materials within ordinary preparation time, the implementation needs attention.
These are suggested evaluation questions, not validated measures or a promise of improvement. They should be adapted with relevant educational expertise.
Keep the context attached to the example
National initiatives operate within particular curricula, governance arrangements, languages and support systems. A school network cannot import those conditions simply by purchasing a similar service.
Ask which authority makes the decision in the original setting and which authority would make it locally. Identify the staff time, expertise and continuing resources involved. Consider learners who need another format or a different route to the same educational objective.
These questions are especially important when an example comes from upper-secondary education and the proposed local users are younger. A change of cohort is a material change, not a minor adaptation.
Our guide to turning AI policy into classroom practice explores the school-level work of connecting principles with everyday decisions. International comparison can inform that work; it cannot replace it.
Make the education conference a comparison exercise
An institutional event can use these cases to improve the quality of discussion.
Give participants a short account of each initiative with its source, date, intended audience and implementation stage. Ask them to identify which statements describe requirements, design intentions, reported activity or demonstrated effects. Leave unanswered questions visible.
Then invite each group to name one local problem and one specific idea it might investigate. They should explain what would need to change in their own setting and what evidence they would seek before extending the proposal.
A keynote can establish these distinctions. A working session allows teachers, leaders and specialists to test them against actual responsibilities. Neither format should present an international example as a ready-made answer.
Our guide to choosing an AI in education keynote speaker helps organisers connect an event with the judgments participants need to make.
What this comparison can support
This article reviews selected official web resources, including a programme profile, French guidance and a service-note subsection, and a Korean-language ministry announcement. It does not review every relevant national policy, the full Estonian evaluation report or the complete history of Korean digital-textbook regulation.
Official sources are useful for identifying stated designs and requirements. They are not independent evaluations of their own educational effects. Future stages remain future stages, and a publication's ambition should remain distinguishable from what has been established.
Learning Coach's position is that international examples become valuable when they sharpen a local educational question. Before asking which country to follow, ask which intervention you understand well enough to examine—and what would persuade you that it helps your learners.
Planning an education leadership event on international AI policy and learning? Explore Learning Coach's international keynotes and request a tailored proposal.
References
e-Estonia, AI Leap. Official programme profile; undated page, accessed 29 October 2026.
French Ministry of Education, Cadre d'usage de l'IA en éducation. Ministry web summary, 13 June 2025.
DGESCO/DNE, digital competence training, assessment and certification arrangements. Service note of 23 January 2026, published 5 February; Pix AI subsection.
Korean Ministry of Education, teacher-authored AI textbooks and educational materials. Official communication, 18 September 2026; English descriptive title. Sources checked 29 October 2026.
