AI literacy: what the EU AI Act asks of your primary school

22 November 2026 · 6 min read · W Consultants

Article 4 of the EU AI Act also applies to school boards. What the duty means since July 2026, what is banned or high risk in education and how to record your team's training.

In short
  • If teachers use ChatGPT, Copilot or an AI feature in a teaching scheme, the school board is a deployer under the EU AI Act.
  • The board must take measures that support the AI literacy of its staff. Since 27 July 2026 this is set out in an amended Article 4.
  • Emotion recognition is banned in education. AI that evaluates learning outcomes or decides a pupil's level counts as high risk.

Even if nobody officially chose AI

The EU AI Act distinguishes between providers, who build AI systems, and deployers, organisations that use AI in their work. For a Dutch primary school, that is the school board. A teacher who has a weekly update drafted, a SENCO who has a protocol summarised or a head teacher who has a staffing plan worked out: it all happens under the board's responsibility.

That also applies if the school never chose AI. A teacher working in the evening with a private ChatGPT account does so for school, and sometimes with information about pupils.

What Article 4 requires

Article 4 is about AI literacy: the knowledge and skills to use AI sensibly, see the opportunities and risks, and know what can go wrong. The duty has applied since 2 February 2025. The Digital Omnibus (Regulation (EU) 2026/1744) amended the text. Since 27 July 2026, providers and deployers must take measures to support the development of AI literacy of their staff.

A fixed level is no longer required, but you must be able to show you took it seriously. Article 4 has no fine of its own. After an incident, a complaint from parents or a question from the inspectorate, people will want to see what the school did.

This is a summary, not legal advice. If in doubt, check the current text or ask your board's data protection officer.

What is banned or high risk in education

  • Banned since February 2025: AI that reads the emotions of pupils or staff, and AI that manipulates people in a vulnerable position, such as children.
  • High risk: AI that evaluates learning outcomes, decides which level of education suits a pupil, or monitors pupils during tests. Strict requirements apply. A teacher does not switch such tools on alone; that is a board decision, with the participation council (MR).

Who needs which training?

  • Teachers and teaching assistants: what AI can and cannot do, never entering traceable pupil data, using AI for lesson preparation and parent contact, and teaching pupils to use chatbots critically.
  • SENCOs (intern begeleiders): in addition, the line with group plans, support plans and safeguarding concerns, and why AI never gives a diagnosis or secondary school advice.
  • School leaders and boards: AI policy, agreements with the MR, the privacy covenant with suppliers, and recording who completed what.
  • Administration: working with documents and figures without pupil or staff data ending up in a chat.

How to record it

  1. Ask who uses AI and for what, including supply teachers and teaching assistants.
  2. Write a short AI policy: which tools and accounts are allowed, which data never, and who to ask.
  3. Choose training per role, for example on a training day.
  4. Record who completed what, with a date. An overview of progress and certificates per team member is exactly that evidence.
  5. Review it every year, with the MR too. Tools change fast.

Read more about the line with pupil data in ChatGPT and pupil data: what is and is not allowed.

More articles