AI Ban Primary School Rules: Netherlands, Norway, NYC Compared

In the span of five months, three very different places reached the same conclusion about young children and chatbots. Norway banned generative AI for students in grades 1 through 7 starting this school year. New York City paused all student-facing generative AI through 8th grade. And on October 30, 2026, a Dutch parliamentary motion backed by parties holding 88 of 150 seats asked for a formal AI ban in primary school. The AI ban primary school debate is no longer hypothetical. But the three policies are not identical, and the differences tell parents more than the headlines do. None of them, notably, bans learning about AI.

Key Takeaways

• Norway banned generative AI tools for grades 1–7 (ages 6 to 13) from late August 2026; ages 14–16 may use it only with teacher supervision.
• New York City imposed a one-year moratorium on student-facing generative AI through 8th grade, banned companion chatbots in all grades, and added twice-yearly AI literacy classes for high schoolers.
• In the Netherlands, a motion from Volt and CDA has majority support, but it is a request to the cabinet, not yet a rule, and new Dutch learning goals from 2027 still require pupils to learn about AI.
• The shared logic: AI can let a child skip the effortful steps that build reading, writing and math, and young children tend to trust a confident machine.
• The evidence on bans themselves is thin; the evidence on unguarded AI hurting learning is stronger. Parents can apply the logic at home without waiting for a law.

What each place actually decided

An AI ban in primary school is a rule that keeps generative AI tools, such as ChatGPT-style chatbots, out of young students’ hands during schoolwork. The three versions in the news differ in scope, age range and legal status.

Norway. According to Engadget, citing Reuters, Norway is banning generative AI for students in grades 1 through 7, roughly ages 6 to 13, from the school year that began in late August 2026. Students aged 14 to 16 can use it only under a teacher’s supervision, and those 17 and older are encouraged to use it appropriately on their own. Prime Minister Jonas Gahr Støre argued that AI can let children skip crucial steps, and that schools should focus on teaching them to “read, write and do mathematics.” It follows Norway’s earlier move to take phones out of classrooms.

New York City. On September 2, 2026, Mayor Zohran Mamdani announced a one-year moratorium covering “all software that uses student-facing generative AI” for students through 8th grade, as the Associated Press reported. The mayor’s office called it the most expansive prohibition in the country. Companion chatbots are banned in every grade. High schoolers get AI literacy classes twice a year, and some will join supervised pilots. Teachers can still use AI for planning. The city also recommended screen-time caps of 30 minutes a day for grades 3–5 and 45 minutes for grades 6–8. We covered the details in our guide to the NYC pre-K to 8 AI ban.

The Netherlands. On October 30, 2026, DutchNews reported, citing broadcaster NOS, that Volt and the Christian Democrats (CDA) had filed a motion asking the cabinet to create a formal rule barring primary pupils from tools like ChatGPT, with 10 parties holding 88 of 150 seats in support. A motion is a request, not a law. And the new primary learning goals taking effect in 2027 still require pupils to explore AI and learn to “interact responsibly with an AI system.” The motion acknowledges that and doesn’t rule out teaching about AI.

Three places, three reasons

Read the statements closely and each place emphasizes something different.

	Norway	New York City	Netherlands
Status	In force since late Aug 2026	One-year moratorium, 2026–27	Parliamentary motion; no rule yet
Ages covered	Grades 1–7 (6–13); supervised use 14–16	Pre-K through grade 8	Primary school (roughly 4–12)
Main stated reason	Children skipping essential steps in reading, writing, math	Teach kids to question; evaluate tools before use	AI “widening learning gaps”; basic skills
Learning about AI?	Older teens encouraged to use it well	AI literacy classes in high school	2027 goals still require AI education
Companion chatbots	Not addressed in coverage	Banned in all grades	Not addressed in coverage
Main critic’s point	Coverage cites no organized opposition	Fairplay: one year is “not nearly enough”; UFT: questions unanswered	PO-Raad: a total ban came “out of the blue”

Norway’s reason is skill-building. Støre’s framing is about the developmental sequence: a child who has a machine write the paragraph never gets the practice of writing it.

New York’s reason is caution and evaluation. The moratorium is explicitly part of “a broader assessment” and comes with a review of all tech tools to cut those “nonessential to learning.” Mamdani framed it as modeling skepticism: “we teach them to ask questions, to challenge assumptions, to interrogate the premise, to ask why. When it comes to AI in our schools, we hold an obligation to do the same.” Critics came from both sides. Josh Golin of the advocacy group Fairplay said “a year to evaluate is not nearly enough time.” The teachers’ union president said the policy “leaves many questions unanswered.”

The Dutch reason is equity. Volt leader Laurens Dassen said: “I believe in the opportunities AI offers us, but at the same time generative AI is widening learning gaps.” The primary schools council, the PO-Raad, agreed that early research shows pupils learn less when they use generative AI heavily, but objected to a total ban arriving without negotiation. A teacher quoted by NOS worried that ruling AI out entirely risks children learning too little about it.

Why an AI ban primary school policy targets young kids: the research

The reasoning behind these bans rests on two mechanisms. Both have research behind them, though neither has been tested on 7-year-olds in the way policymakers might wish.

Mechanism one: skipping the productive struggle. The best evidence comes from older students. In a field experiment with nearly 1,000 high school math students in Turkey, published in PNAS in 2025, Hamsa Bastani and colleagues gave some classes a plain GPT-4 chat interface and others a version designed to give teacher-written hints instead of answers. With the plain chatbot, practice scores rose 48%, but on a later exam without AI those students scored 17% lower than classmates who never had it. The guarded tutor largely avoided that drop. The researchers’ explanation: students used the plain chatbot as a “crutch.” The most common message they sent was “What is the answer?”

That is the core engineering problem. A general chatbot is optimized to complete the user’s request. A child’s request is usually “make this homework done.” Learning requires the opposite: the child doing the hard part.

The OECD’s Digital Education Outlook 2026, summarized by its education director Andreas Schleicher, makes the same point at a system level: when AI “removes the productive struggle essential for learning,” students may finish faster with better immediate results, “but their understanding may be less deeply consolidated.”

Mechanism two: young children trust confident machines. In a January 2026 OECD essay on designing safe AI for education, Schleicher put the age argument plainly: “A six-year-old and a sixteen-year-old may tap the same screen, but they are not engaging with the same reality.” For younger learners, he wrote, AI “must behave less like an oracle and more like a guardrail,” with no pretending to be a friend, because “children are wired to trust authority. If AI speaks with confidence, they will believe it.” Bastani’s team found the plain chatbot gave a correct answer only about half the time on their problems. An adult might catch that. A third-grader mostly won’t.

Schleicher’s analogy is a useful one for parents: “Scissors in the early grades are blunt. Chemistry labs are locked until students are trained.” Graduated access, not permanent prohibition.

What the evidence doesn’t tell us yet

Honesty requires a few caveats here.

There is no published study yet of what Norway’s or New York’s ban does to learning outcomes. They are months old. Even phone bans, the closest precedent, are still being evaluated years later, so nobody should expect quick, clean answers about test scores.

The studies on AI harming learning mostly involve teenagers and college students, not 6- to 10-year-olds. Extending them downward is reasonable, since younger kids have less background knowledge to judge AI answers, but it is an inference.

And the pro-AI case has evidence too. Well-designed tutors that withhold answers can help, as Bastani’s guarded version and other trials show. The OECD Outlook itself says AI tools “purpose built for learning show more promise.” The bans target general-purpose chatbots handed to young children, which is where the evidence of harm is clearest.

UNESCO’s 2023 guidance on generative AI in education already recommended setting an age limit for independent conversations with these platforms. These policies are governments catching up to that advice.

What parents can take from this at home

Separate “learning about AI” from “using AI to do the work”

All three places draw this line. The Dutch motion allows AI education; New York teaches literacy; Norway lets older teens use it well. At home, you can explain how a chatbot works, what it gets wrong and why, without letting it write a 9-year-old’s book report. Our article on what happened when schools banned or integrated ChatGPT explores the trade-off.

Use age, not a calendar, as your guide

Norway’s three tiers are a sensible template: no independent use for younger kids, supervised use for early teens, guided independence for older ones. You can adapt the ages. The principle is that access grows with the skills to judge the output.

When AI is used, make it ask, not answer

If your younger child encounters AI, at school or a relative’s house, the design that research supports is hint-giving, not answer-giving. You can set that up yourself: sit beside them and have the AI quiz them, or ask it to explain a step without giving the result. Then close the laptop and have them try one alone.

Protect the basic skill time

Støre’s “read, write and do mathematics” is a reasonable home priority, ban or no ban. Handwritten practice, reading aloud and mental math do not need a policy to be worthwhile.

What not to do: treat a ban as the end of the conversation

A school ban covers school devices and school hours. It doesn’t cover the phone at home, the chatbot built into a messaging app, or an older sibling’s laptop. Kids in banned districts still need to understand AI, and parents still need to set home rules. Compare the experience in Sweden’s and Norway’s earlier screen reversals.

What to Watch For Over the Next 3 Months

• Week 4: Find out your own school’s policy. Does it distinguish between teachers using AI, students using AI, and students learning about AI? If you can’t find it, ask.
• Month 2 red flags: Your young child using a chatbot to finish assignments at home and being unable to do a similar problem without it, or treating the chatbot’s answers as always right. Both signal that the “skip the steps” risk is real for your kid.
• Month 3 self-check: Watch for the Dutch cabinet’s response to the motion, early reports from New York’s moratorium year, and any Norwegian evaluation. Ask yourself: does your home have an age-based rule, and does your child know why it exists?

Frequently Asked Questions

Has the Netherlands banned AI in primary schools?

Not yet. On October 30, 2026, a motion from Volt and CDA asking the cabinet to create a formal ban had the support of 10 parties holding 88 of 150 seats, according to NOS. A motion is a request to the government. The cabinet still has to act, and new learning goals from 2027 still require pupils to learn about AI.

What ages does Norway’s AI ban cover?

Norway’s ban covers grades 1 through 7, about ages 6 to 13, from the school year that started in late August 2026. Students aged 14 to 16 may use generative AI only with teacher supervision. Those 17 and older are encouraged to use it appropriately on their own, according to Reuters reporting summarized by Engadget.

Is there evidence that banning AI for young kids improves learning?

Not yet; the bans are too new to have been evaluated. The evidence supporting them is indirect: studies like Bastani et al. (PNAS, 2025) show unguarded chatbots can raise practice scores while lowering later exam performance. Most of that research involves teenagers, so applying it to young children is a reasonable inference, not a proven result.

Does an AI ban mean my child won’t learn about AI?

Not in these three cases. New York added AI literacy classes for high schoolers, the Dutch motion leaves room for teaching about AI, and Norway encourages older teens to use it well. The bans target young children using generative AI to do schoolwork, not education about how AI works.

Should I ban AI at home for my elementary schooler?

That’s a family decision. The policies above suggest a middle path: no independent chatbot use for young kids, supervised use where AI asks questions instead of giving answers, and plenty of time on reading, writing and math done by hand. Explain your reasons so the rule teaches something.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. DutchNews.nl. (2026, October 30). “MPs call for ban on generative AI in primary schools.” https://www.dutchnews.nl/2026/10/mps-call-for-ban-on-generative-ai-in-primary-schools/
2. Izaguirre, A. (2026, September 2). “NYC, U.S. largest school system, bans AI for students through 8th grade.” The Associated Press, via CP24. https://www.cp24.com/news/world/2026/09/02/nyc-us-largest-school-system-bans-ai-for-students-through-8th-grade/
3. Engadget. (2026, June). “Norway imposes broad restrictions on AI for elementary school kids” (citing Reuters). https://www.engadget.com/2198117/norway-imposes-broad-restrictions-on-ai-for-elementary-school-kids/
4. Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). “Generative AI without guardrails can harm learning: Evidence from high school mathematics.” Proceedings of the National Academy of Sciences, 122(26). https://doi.org/10.1073/pnas.2422633122
5. Schleicher, A. (2026, January 19). “How to effectively use Generative AI in education” (on the OECD Digital Education Outlook 2026). OECD Education and Skills Today. https://oecdedutoday.com/how-to-effectively-use-generative-ai-in-education/
6. Schleicher, A. (2026, January 23). “Designing safe AI systems for education.” OECD Education and Skills Today. https://oecdedutoday.com/designing-safe-ai-systems-for-education/
7. UNESCO. (2023). “Guidance for generative AI in education and research.” https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research