Teach17

Norway has taken one of Europe’s strongest positions on children using generative artificial intelligence in school. From autumn 2026, pupils in Years 1–7 should generally not be given access to AI tools for their schoolwork.

Despite some dramatic headlines, Norway has not banned every form of artificial intelligence from every classroom. Its approach is more thoughtful: establish secure foundations first, introduce AI gradually and expect older students to learn how to use it critically.

Here are 17 things teachers and school leaders should know.

1. It is not quite a complete ban

Norway’s government announced a new national recommendation on 19 June 2026. It states that pupils in Years 1–7 should, for the most part, not be given access to artificial intelligence.

The words “for the most part” are important. This is not a blanket prohibition covering every pupil, teacher, subject and AI-powered tool. It is a strong expectation that young children should not routinely use generative AI to complete schoolwork.

2. Norway is introducing AI according to age

The recommendations divide pupils into three broad stages.

Pupils in Years 1–7 should generally not use AI themselves. In Years 8–10, it can be introduced gradually and under teacher supervision. In upper secondary education, students should learn to use AI critically and appropriately in preparation for further study and employment.

This is better understood as delayed and controlled access rather than an attempt to remove AI from education permanently.

3. Children should learn before they automate

Norway’s central argument is simple: pupils should learn to read, write, calculate and think before they begin using technology that can perform these processes for them.

Writing an explanation, organising an argument and struggling through a calculation are not merely ways of displaying learning. They are part of how learning happens.

When AI completes these stages, a pupil may produce an impressive answer without developing the knowledge or thinking behind it.

4. A better answer does not always mean better learning

Generative AI can quickly improve the apparent quality of a pupil’s work. However, the OECD’s 2026 Digital Education Outlook warns that improved task performance does not automatically produce lasting learning.

Some studies found that pupils using general-purpose AI produced better work while the tool was available, but the advantage disappeared—and sometimes reversed—when they later completed examinations without it.

Schools therefore need to separate two questions:

Did the pupil produce a good answer?

Did the pupil actually learn how to produce that answer?

5. Productive struggle still matters

Learning often requires pupils to retrieve information, make mistakes, reconsider an idea and try again.

The OECD warns that poorly designed AI use can remove this productive struggle. Pupils may finish work more quickly while developing less cognitive stamina, perseverance and deep understanding.

The fastest route to a finished task is not always the best route to long-term learning.

6. Norway is responding to falling foundational skills

The decision is part of a wider concern about declining results in Norwegian schools.

The Norwegian government says that international assessments, including PISA and PIRLS, have shown significant falls in reading and other foundational skills. It reports that around one-quarter of Norwegian pupils read below the OECD’s minimum proficiency threshold.

This does not mean generative AI caused the decline. The technology is too recent for that conclusion. However, Norway does not want its rapid adoption to weaken foundational learning further.

7. AI had already spread rapidly through schools

The policy has not been introduced before pupils encountered AI. Its use was already increasing quickly.

A January 2026 government announcement reported pupil use of AI in nearly three-quarters of compulsory schools and more than 90% of upper secondary schools. Among surveyed leaders, 65% in primary schools and 89% in lower secondary schools said pupils were using AI.

Norway is therefore attempting to regain control of a technology that entered classrooms faster than consistent policies could be developed.

8. Many schools were making their own rules

Before the national announcement, there was no recommended nationwide age limit for AI use in Norwegian schools.

Some schools had introduced local restrictions, while others left decisions to individual teachers, pupils or parents. Around half of schools containing primary-aged pupils said they had introduced, or were introducing, age limits.

National guidance may reduce the confusion caused when neighbouring schools take completely different approaches.

9. Younger pupils may not recognise confident nonsense

Generative AI can produce fluent, plausible and incorrect information.

Experienced adults may question a suspicious claim, compare several sources or notice that an explanation does not fit their existing knowledge. Younger children have less background knowledge with which to make these judgements.

The danger is not only that AI gives the wrong answer. It is that it can give the wrong answer in language that sounds completely convincing.

10. Pupils can learn about AI without using it for every task

Norway is not suggesting that children should be kept ignorant of artificial intelligence.

Its curriculum guidance says pupils can learn what generative AI is, how it works, why it has limitations and how it affects individuals and society. Schools can also teach source criticism, copyright, digital judgement, ethics and responsible use.

A pupil does not need unrestricted access to a chatbot to learn that chatbots exist.

Teachers might demonstrate an AI response to a class, identify its errors and discuss how it could be improved without asking every pupil to generate their own work.

11. Teacher knowledge should come before pupil access

Norway says that AI can be introduced in Years 8–10 only when teachers have developed sufficient knowledge and judge that its use will support learning.

This reverses a pattern often seen with educational technology: pupils gain access first, while teachers are expected to work out the rules afterwards.

Schools should establish staff training, safeguarding procedures and clear educational purposes before rolling out a new AI platform.

12. Teacher use and pupil use are not identical

Norway’s education minister has distinguished between a teacher using AI to support planning and a pupil using it to avoid thinking.

A knowledgeable teacher might ask AI to suggest quiz questions, simplify an initial explanation or produce ideas for a lesson. The teacher can then check, edit or reject the result using professional judgement.

A young pupil may lack the knowledge needed to recognise an inaccurate, biased or poorly designed response.

Teacher use still needs rules, but it should not automatically be treated as equivalent to pupil use.

13. Some pupils will still need AI-based support

Norway’s recommendations allow for exceptions.

AI-based tools may support pupils receiving adapted teaching, specialist language support or help connected to additional needs. The Norwegian Directorate for Education is developing further guidance about when these exceptions should apply.

A sensible restriction should prevent AI from replacing learning without removing technology that gives a pupil meaningful access to learning.

14. The policy is part of a wider digital reset

Norway’s AI decision is not an isolated announcement.

The government has linked it to a wider programme involving less uncritical screen use, greater access to physical textbooks and changes to the curriculum. Its stated aim is to place more emphasis on books, reading, writing, mathematics and clearly defined knowledge.

The principle is not that digital technology is always harmful. It is that technology should earn its place by improving learning.

15. Restricting school access will not stop home use

One weakness of the policy is that pupils can still access generative AI outside school.

This could create a new divide. Some children may receive careful guidance from informed parents, while others use public chatbots without supervision. Some may submit AI-generated homework even though they cannot access the same tools in lessons.

Schools will still need to teach honesty, attribution and responsible AI use. A school restriction alone cannot control what happens on personal devices at home.

16. Schools will struggle to define what counts as AI

It may be relatively easy to block a well-known chatbot. It is harder to identify every AI feature embedded within search engines, word processors, translation software, revision platforms and accessibility tools.

Schools will need policies based on what the technology is doing, rather than a list of prohibited brand names.

A useful distinction might be:

• AI that supports access;
• AI that gives feedback;
• AI that suggests ideas;
• AI that completes the thinking;
• AI that produces work presented as the pupil’s own.

These uses do not create the same educational risks.

17. The strongest lesson is to use AI deliberately, not automatically

Norway’s approach is unlikely to provide the perfect solution for every education system. A restriction that is too rigid could leave pupils unprepared for later study and employment. An approach that is too relaxed could allow AI to replace the reading, writing, recall and problem-solving pupils most need to practise.

The most useful principle is gradual release.

Schools could:

• protect foundational knowledge and independent practice;
• teach pupils about AI before expecting them to use it;
• introduce tools at an age-appropriate stage;
• identify tasks where AI would remove the intended learning;
• preserve appropriate support for pupils with additional needs;
• train teachers before introducing new platforms;
• require older pupils to question, verify and acknowledge AI output;
• redesign assessment so that teachers can still see what pupils know.

The OECD concludes that generative AI is most valuable when it is used for a clear educational purpose and does not replace cognitive effort or teacher judgement.

Norway’s message is not simply that AI is bad. It is that children should develop something worth augmenting before schools ask artificial intelligence to augment it.