◆ Norway to control classroom AI access in three age-based stages
The Norwegian government will enforce a policy to regulate the use of generative artificial intelligence (AI) in elementary and middle school classrooms.
Norwegian Prime Minister Jonas Gahr Støre announced at a press conference last month that the government will, in principle, ban the use of generative AI for students from first grade through seventh grade, ages 6 to 13. The policy will take effect in late August, when the new school year begins.
Generative AI refers to programs that produce text, images, or calculation results when a user inputs a question. ChatGPT is a representative example. The issue of students outsourcing homework or problem-solving to such programs has become a major point of contention in education.
The policy is divided into three stages by age. Students ages 6 to 13 will be prohibited from using generative AI at school. Students ages 14 to 16 will be allowed to use it only under direct teacher supervision, and teachers must first complete related training. Students 17 and older will be encouraged to use AI responsibly in preparation for college and employment.
In his announcement, Støre said, “The most important thing in school is for children to learn how to read, write, and do arithmetic.” The Norwegian government believes AI use could cause younger students to skip important stages in their education.
The move is seen as an extension of a policy line Norway has pursued for years. In 2024, Norway banned smartphone use in schools. The basis for the debate at the time was a doctoral dissertation by Sarah Abrahamson, a researcher at the Norwegian School of Economics (NHH), which found that students’ grades and mental health indicators improved at schools that collected phones. Along with the latest announcement, the government also proposed increasing the budget to expand the use of printed books in classrooms instead of tablets.
Two months before this announcement, the Norwegian government also unveiled legislation to set the minimum age for social media membership at 16. The move followed Australia’s landmark ban last December on social media use by people under 16, the first of its kind in the world. Analysts say a trend is spreading across Europe and Oceania in which governments are increasingly managing children’s digital environments.
Norway is regarded as the first country to restrict access to generative AI for young students on this scale. Because the decision comes as education authorities around the world review how AI should be introduced into classrooms, it is expected that the success or failure of Norway’s policy will influence other countries’ judgments.
◆ PISA math scores at record low… Brookings says the risks outweigh the benefits
The direct trigger for the policy is declining academic performance. Norwegian students’ PISA math scores have fallen to their lowest level on record.
PISA is an assessment by the Organisation for Economic Co-operation and Development (OECD) that compares the academic achievement of 15-year-old students across countries, and it is used as an international benchmark for education levels. Norway has also shown a downward trend in reading.
Norway is known for its strong digital infrastructure and for being an early adopter of new technologies. That such a country concluded that access to AI before foundational learning is established is harmful to education has led observers to predict that the decision may influence policy discussions in other countries.
Norway is restricting access to generative AI in elementary and middle school classrooms in stages, citing declining foundational skills and concerns that the learning process is being bypassed. (Photo=Solution News Magnipic)
There is also research to support the decision. In a report released earlier this year, the Center for Universal Education at the Brookings Institution concluded that, at present, the risks of using generative AI in children’s education outweigh the benefits. The researchers interviewed more than 500 students, teachers, and parents in 50 countries and reviewed more than 400 studies.
The report identified skipping the learning process as the biggest risk. When students write or solve problems, the thinking process itself is part of how they build skills, but if AI produces the final output instead, that process disappears. For adults who already have foundational skills, AI can improve productivity; for children still developing those skills, it can mean missing a key stage of development, the report said.
There have been past examples of tools replacing human abilities. Calculators replaced mental arithmetic, and keyboards replaced handwriting. But the report noted that generative AI differs from earlier tools in that what it replaces is not an individual function but the entire thinking process. It explained that this is similar to the difference between an adult who has learned multiplication tables using a calculator and a child who depends on a calculator without having learned them.
Rebecca Winthrop, director of the Center for Universal Education at Brookings, warned that children who rely on generative AI to provide answers stop thinking for themselves. They do not learn how to distinguish fact from falsehood, judge strong arguments, or understand different perspectives. Winthrop described this phenomenon as “cognitive underdevelopment,” expressing concern that a pattern similar to cognitive decline in older adults could appear in young learners.
The report also pointed out that AI dependence creates a repeating cycle. The more students leave thinking to AI, the weaker their own thinking abilities become, and the weaker those abilities are, the more they depend on AI again.
There are examples in the opposite direction as well. The United Arab Emirates has made AI education mandatory starting in kindergarten. Germany considered a ban but did not adopt one. The Brookings report also stated that AI combined with teaching methods that are designed and verified for educational purposes can help learning. What the report criticized was the current use of general-purpose AI, developed for adults, entering classrooms without rules or safeguards.
◆ South Korea’s AI digital textbooks shift to voluntary choice one year after launch
The Norwegian government is changing the direction of education policy so that young students build reading, writing, and thinking skills first rather than relying on AI. (Photo=Solution News Magnipic)
South Korea’s situation is also linked to Norway’s decision. The South Korean government introduced AI digital textbooks in 2025.
The first subjects were math, English, and information technology for third- and fourth-grade elementary students, first-year middle school students, and first-year high school students. The stated goal was customized education that provides learning content tailored to each student’s level.
The policy direction changed within a year. In its 2026 work report, the Ministry of Education used the term “AI educational materials” instead of “AI digital textbooks.” The implementation method also shifted from universal adoption to school autonomy.
Educators say this is less a refinement of policy than a phase of adjustment after a rushed rollout. Some also argue that the single-textbook-centered design failed to function properly in the field.
Concerns were raised from the start. Lawmakers pointed out that young elementary students were not adept at operating devices, and the National Association of Superintendents of Education conveyed concerns about budget shortages and privacy protection to the ministry.
Analysts say the difference between Norway and South Korea is also visible in the conditions for teacher preparation. Norway required teachers to complete training before allowing AI use for students ages 14 to 16. South Korea, by contrast, proceeded with teacher training after setting the rollout schedule. The two countries appear to have reversed the order of technology introduction and field preparation.
Data management for learning records remains another issue. The AI textbook project involved a budget measured in trillions of won. The number of elementary, middle, and high school students in South Korea in 2026 is estimated at about 4.83 million, and their learning records will be accumulated, but critics say no dedicated protection system has been established. The structure records which questions a student got wrong and where they spent a long time working.
It also differs from global regulatory trends, analysts say. The United States has revised its rules to restrict the commercial use of children’s data. The European Union classifies educational AI as high-risk and applies separate regulations. Critics point to the contrast with South Korea, where the market was expanded first without dedicated protections for children’s learning data.
In its 2026 work report, the Ministry of Education placed foundational learning at the center of policy. The direction is to limit AI to a learning aid and define basic academic skills as a state responsibility.
◆ The solution lies in designing the sequence of introduction… “AI should come after foundational skills are formed”
The solution suggested by the Norwegian case and the Brookings report comes down to designing the timing and order of AI adoption. The researchers concluded that development has a sequence, and if reading, writing, and arithmetic are bypassed during the period when those abilities are being formed, it is difficult to recover later.
Norway’s three-stage age-based system is seen as an application of this principle in policy form. During the ages of 6 to 13, when foundational academic skills are being formed, students are made to go through the thinking process themselves, and only after those abilities are established do they use AI as a tool.
This is interpreted not as a policy of excluding AI, but as one of adjusting the order of introduction. The design, which broadens access as students move to higher grades, also reflects the premise that AI literacy itself is necessary.
Observers say this is the key implication for South Korea as well. The issue is not whether AI should be excluded from classrooms, but how the state should design its introduction at each developmental stage.
Norway has set that standard by age, while South Korea is reorganizing policy around foundational learning. The approaches differ, but educators say the shared conclusion is that foundational development comes first.
The Brookings report proposed three lines of action: develop AI tools that support validated teaching methods, provide AI literacy education for both students and teachers, and build design norms and management systems that prioritize safety, privacy, and healthy development. The proposal is to shift from AI that gives answers to AI that helps students reach answers on their own.
AI literacy refers to the ability to verify, select, and use the results produced by AI. The report notes that this ability also develops on top of the foundations of reading, writing, and arithmetic. Without those foundations, there is no basis for judging whether AI output is right or wrong.
Emotional development was also identified as a challenge. In one survey, 65 percent of girls ages 5 to 13 said they regarded AI as a friend. Among those ages 11 to 13, 51 percent said they had turned to AI for help when they felt sad or anxious. There are concerns that if technology is introduced ahead of teacher training and parental involvement, not only learning but also emotional development could be skipped.
Establishing a system to verify effectiveness was also raised as a task. Winthrop suggested developing indicators to track and measure children’s cognitive development in relation to AI use so that it can be determined whether the tool helps or harms learning. She also said that AI used in a narrow way for a specific purpose can support thinking and learning.
The Brookings Institution said the damage done so far is significant but recoverable. Norway’s policy will take effect in August, and education authorities around the world are expected to watch the results closely.
The solution lies in designing the sequence of introduction… “AI should come after foundational skills are formed”