At what age can a child use AI?

There is no scientific consensus on a "safe" age. On the private rules side, most consumer chatbots (ChatGPT, Gemini, Copilot) require at least 13 years old in their terms of use, and a parent's agreement before 18. Norway sets a school benchmark instead: no generative AI in class before year 7 (around 13 years old), then supervised use up to 16.

Is ChatGPT allowed in schools in France?

France is not following the Norwegian route. In 2025, the Ministry of Education published a "framework for use" of AI in education that is broadly favourable, with tools offered to volunteer teachers (the France 2030 programme). So there is no national ban comparable to Norway's: in practice, use depends on the rules set by each school and each teacher.

How should AI use for homework be framed?

A few concrete markers, since school does not cover the home: favour "tutor" use (having a concept explained, checking a line of reasoning) over "ready-made answer"; ask the child to produce something first, then compare with the AI; check the sources, because AI regularly invents things (hallucinations); and keep AI-free stretches to maintain the habit of effort. The aim is not to ban, but to learn to use it with some distance.

Can AI help a child learn?

Yes, under conditions. As a personalised tutor, it can re-explain a concept endlessly, adapt to the pupil's pace, drill a language or support a struggling child — that is the bet countries like Estonia are making. The benefit depends, however, on two things: a tool designed to make pupils think (rather than answer in their place), and the presence of an adult. Badly used, the same AI replaces effort instead of supporting it.

Does AI make pupils less intelligent?

The honest answer: it depends on the use, and the science is still young. Used as an answering machine, AI encourages "cognitive offloading" — you delegate the effort, you retain and understand less. Used as a tutor that explains, it can help instead. No study so far proves a lasting drop in intelligence: the available work mainly concerns adults and remains preliminary.

Did Norway ban AI in schools? What the rule really says

When a digitally rich country unplugs AI for its youngest

Norway is not a technophobic country. It is one of the most connected in Europe, a pioneer of the digital school. That is precisely what makes its decision striking: on 19 June 2026, Prime Minister Jonas Gahr Store announced a near-total ban on generative AI in primary school, applicable from the start of term at the end of August. Not a local experiment: a national rule. The news travelled the world within hours — often distorted on the way. Let us untangle it, without hype: what is true, what it is worth, and what it will not fix.

What Norway actually decided

First, correct two shortcuts doing the rounds. This is not “AI banned for under-12s”, and it is not a law that would follow the child into their bedroom. The measure governs AI within the school setting, in age tiers:

Years 1 to 7 (around 6-13 years old): near-total ban on generative AI — chatbots, image generators, homework assistants.
14-16 years old: use allowed, but under a teacher’s supervision.
17 and over: independent use encouraged, “where appropriate”.

Three clarifications that change the reading. One: these are national standards imposed on schools, not a criminal ban aimed at families. Two: the measure says nothing about use at home — a child can perfectly well use a chatbot for homework in the evening. Three: Norway is not rejecting AI on principle. The proof is that these are tiers, not a blanket ban: older pupils have access, first under supervision, then freely. The message is not “never any AI”, but “not too early, and not without an adult”.

Why now: Store’s argument

The official justification is pedagogical, not neurological. For Store, the most important thing at school remains that children learn to read, write and count; AI would risk making them “skip crucial stages” of that learning. That is all. The idea of a “child’s brain that would not work as it should” — widely shared on social media — belongs to media commentary, not to government statements. That is exactly the kind of drift to watch for: a sober decision dressed up as a brain alert.

Norway is in fact applying a recipe it believes it has already validated: in 2024, it took smartphones out of classrooms. AI follows the same logic — protecting the fundamentals from a technology judged too much of an “accelerator”.

The scientific argument… and its limits

This is where Le Recul insists, because this is where everyone gets carried away.

On AI. The most-brandished study is Your Brain on ChatGPT: Accumulation of Cognitive Debt, published by the MIT Media Lab in June 2025. Three groups write essays — with a chatbot, with a search engine, or with no tool at all. Over four months, the “AI” group shows weaker neural connectivity, a diminished sense of ownership of their text, and one spectacular statistic: the vast majority of AI users are unable to quote a sentence from their own essay. The authors talk about “cognitive debt”. Except that — a detail systematically forgotten — it is a preprint not peer-reviewed, on a small cohort of adults. It documents a real risk; it does not definitively demonstrate it, and above all not in children aged 6 to 13, who were not studied.

On smartphones. The 2024 Norwegian precedent rests on the work of researcher Sara Abrahamsson (more than 400 middle schools): a fall in bullying, better grades — especially among girls, and more so among those from modest backgrounds — and a drop in psychological consultations. Figures that impress. But here too, researchers add caveats: correlation is not causation, many schools did not really “ban” the phone (just silent mode), and some measured effects are modest or contradict each other from one metric to the next. The precedent is encouraging, not conclusive.

Put plainly: Norway’s protective instinct is defensible, but it is advancing on an evidence base more fragile than the headlines suggest.

The flaw nobody is fixing

Even assuming AI is harmful to early learning, the measure has a structural blind spot: it stops at the school gate.

Generative AI is free, available on any connected phone, tablet or computer — and the publishers are pushing it everywhere, with ChatGPT even trying to become an everyday “super-app”. A ban that only applies during class hours therefore changes nothing about what a child does for the rest of the day. Worse: it risks widening inequalities. In families where a parent supervises and explains the tool, the child learns measured use. In the others, they use AI alone, with no framing — exactly what the school wanted to avoid. A school ban with no family policy is a firewall with a wide-open door next to it.

The European split

The most instructive part is that Norway does not embody a consensus. On the same continent, in the same year, the bets are directly opposed:

Estonia is doing the exact opposite. Its AI Leap programme is rolling out AI tools — in partnership with OpenAI and Anthropic — to tens of thousands of secondary pupils and thousands of teachers, on the idea that knowing how to steer AI is a basic skill to be taught early.
Sweden is pulling back from all-digital and reinvesting heavily in paper textbooks, while describing it as “recalibration, not a reversal”.
France published a framework for use in 2025 that is broadly favourable to supervised AI in class, in the wake of France 2030.
The European Union, for its part, is pushing an AI literacy framework to teach pupils to understand and criticise these tools — a regulatory logic found, in a more turbulent form, in the slipping AI Act timetable.

Two philosophies, no definitive proof on either side. That is the whole subject: nobody yet knows who is right, and today’s children are the test cohort.

What this really implies

Beyond Norway, this decision acts as a revealer. It poses a question every parent and every education system will have to settle: should a child be taught to do without AI before being taught to use it? Skipping the effort — searching, getting it wrong, rephrasing — may well mean skipping exactly what builds a brain capable, later on, of judging an AI’s answers rather than swallowing them.

But the other risk exists too: raising a generation foreign to the tool that will shape its labour market. At a time when AI is already redrawing employment — a movement we follow on the layoffs and reconfigurations side — algorithmic illiteracy could cost as much as dependency. The right dosage is not binary; for now it is undiscovered, and it is honest to say so.

Le Recul’s reading

Norway is neither making a blunder nor an act of clear-sightedness: it is making a cautious bet on uncertain science. Protecting the learning of fundamentals is legitimate. But the measure sells better than it holds: the evidence is preliminary, the “brain” argument is a media add-on, and the flaw — use at home — is not plugged. This is not a ban on AI; it is a school moratorium, time to see.

Our position is simple: neither panic nor wonder. The real fault line is not “for or against AI at school”, but “with or without distance”. And on that point, Norway at least has the merit of taking its time — while others charge ahead.

What to take away

What. Near-total ban on generative AI at school for 6-13 year olds, supervised use from 14 to 16, free at 17 and over.
When. From the start of term at the end of August 2026; announced by Prime Minister Store on 19 June 2026.
The limit. The measure does not cover the home — its main blind spot, acknowledged by the media themselves.
The evidence. The MIT preprint (“cognitive debt”) and Abrahamsson’s smartphone study feed the debate, but neither settles it: preliminary, contested, never tested on young children.
The context. While Norway brakes, Estonia accelerates (AI Leap), Sweden recalibrates, France frames. No European consensus.

The figure to remember: 20,000. That is the number of secondary pupils Estonia gave access to AI through its AI Leap programme, launched in 2025 — while Norway, at the following start of term, chooses to take it away from its youngest. Same continent, opposite bets.

Our verdict. AI promises to accelerate everything; Norway chooses to slow down. On paper, that is caution; in practice, it is a school moratorium with a fragile scientific base and a wide-open back door. Worth watching: the first assessments after the 2026 start of term, and whether other European countries follow — or decide the other way.

Norway bans AI in primary school: protecting children, or holding back the inevitable?