Guide · AI training · Art. 4

AI training in an educational institution.

What the EU AI Act requires, who to train, what to cover and how to record it. With a free template to start today. It applies to schools, vocational colleges and universities; the examples come from a school.

Updated 6 November 2026 · 8 min read

In short

The short answer.

  • Yes, it is mandatory. Since 2 February 2025, every school that uses AI in its work must take AI literacy measures for its staff (Art. 4).
  • No level has to be guaranteed. Since the Omnibus came into force on 27 July 2026, proportionate measures that you can show are enough.
  • No certificate or official hours. There is no compulsory syllabus. An internal record is enough.
  • Different for each role. Leadership, teachers, counsellors and office staff need different things.
  • Already supervised. National authorities have overseen Art. 4 since 2 August 2026.
  • Stricter from 2 December 2027 for anyone overseeing a high-risk use, such as assessing students.

The rule

What Article 4 says after the Omnibus.

The duty has applied since 2 February 2025. Regulation (EU) 2026/1744, the Digital Omnibus, reworded it: it used to require ensuring, “to their best extent”, a sufficient level of AI literacy; it now requires taking measures and makes clear that no specific level has to be guaranteed.

“Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.”

Art. 4(1) · as worded by Regulation (EU) 2026/1744

AI literacy means the skills, knowledge and understanding that allow informed use of AI and awareness of its opportunities, its risks and the harm it can cause (Art. 3(56)).

Does it apply to you?

If AI is used in the school, yes.

A school is a “deployer” when it uses an AI system under its authority in its professional activity (Art. 3(4)). That includes a learning platform with AI features, the assistant in the office suite, an automatic marking tool or a general-purpose chatbot teachers use to plan lessons. The size of the school and who runs it make no difference.

The usual blind spot is the tools staff use on their own initiative. Inventory them too: you cannot train people on what you do not know about.

Who to train

Each role, its own training.

Article 4 asks you to consider what each person already knows, how they use AI and who it is used on. In a school or vocational college it is almost always used on children, which calls for more care; at university, mostly on adults. On top of a common base for all staff, each role needs something different:

RoleWhat it needs, beyond the common base
Leadership teamWhich AI the school uses and how risky it is, what is prohibited, what arrives in December 2027 and the criteria for approving a tool.
TeachersUsing AI in their subject without handing over their judgement: checking what it produces, not entering student data into unapproved tools and taking extra care if AI helps to mark or assess.
Counselling and learning supportSpecially protected data such as health or special educational needs; bias; and why no decision about a student should rest on an automated result.
Office and admissions staffFamily and student data in emails, documents and assistants; which tool is approved for each task.
IT or digital coordinatorSettings and permissions for each tool, which data leaves the school, the provider's terms and how to spot new tools.
Data protection officerWhen an AI use may need an impact assessment or a review of the provider contract, and how it fits with the data protection training the DPO already handles (Art. 39(1)(b) GDPR).
Supervisors, canteen, after-school staff and contractorsThe basics of the tool they use on the school's behalf and its rules of use. They are the “other persons” in Article 4.

What about students? Article 4 does not require training them: they are among the people AI is used on, which counts when shaping staff training. Teaching AI to students is a separate, curricular matter.

Step by step

How to organise it in six steps.

  1. Take an inventoryList the school's AI tools, who uses them and what for, including the ones staff use on their own. “Tools” sheet in the template.
  2. Group by roleDefine the school's roles and which tools each uses. There is no need to go person by person.
  3. Set a common baseA short session for all staff: what AI is, what the school uses, what is prohibited and which data is never entered.
  4. Add the specificsWorkshops or guides by role and tool, above all where student data is processed or AI influences decisions about students.
  5. Record itEach measure with its date, audience, tool, content and where the evidence is kept. “Actions” and “Participants” sheets.
  6. Review it every yearAt the start of the school year, when a new tool arrives and when new staff join. Before 2 December 2027, training for anyone overseeing high-risk uses.

What to cover

The minimum worth covering.

The European Commission suggests four blocks. This is how they apply in a school:

1 · Basics

What AI is and which we use

How a generative model works, why it can be confidently wrong, and the list of tools approved in the school.

2 · Role

The school's role

The school uses other people's AI rather than building it: it is a deployer. That sets its duties: use it as instructed, oversee it and inform people.

3 · Risks

The risks of each tool

  • Made-up or outdated answers.
  • Bias when assessing or describing students.
  • Children's personal data leaving the school.
  • Prohibited practices, such as inferring emotions in class.

4 · Context

Tailored to each role

All of the above, applied to what each role does, with its legal and ethical aspects. Reading the instructions for use is usually not enough.

How to show it

What to record and what to keep.

There is no official format. The Commission says an internal record of training and other initiatives is enough. To be useful for leadership, the DPO or an inspection, each measure should answer these questions:

  • What was done: session, course, workshop, guide or rule of use, and its content.
  • For whom: the role and the people who took part.
  • When: the date and, if any, the duration.
  • Which tool: or whether it was general training.
  • Where the evidence is: sign-in sheet, invitation, materials or platform record.
  • When it is reviewed: the next planned date.

There is no need to test staff or issue certificates.

Free template

AI literacy record for educational institutions.

An Excel workbook ready to fill in. It also works in Google Sheets, Numbers and LibreOffice.

  • Tools: the inventory, with who uses each tool, whether it processes student data, whether it is high-risk and whether it is approved.
  • Actions: each measure, with type, role, tool, content and evidence.
  • Participants: who attended each action.
  • Summary: calculated automatically, by role.

It includes example rows you can delete. The participants sheet holds staff personal data: keep it with restricted access.

Common mistakes

What to avoid.

  • Believing a certificate or a course of so many hours is needed. The AI Act does not require it.
  • Treating training as done once the instructions for use are sent out. The Commission considers that usually insufficient.
  • Training only teachers and forgetting office staff, counsellors or external staff.
  • Running one session and never returning to it. Tools change every year.
  • Training on AI in general rather than on the tools the school actually uses.
  • Training without recording it. What is not written down is hard to show.

Frequently asked questions

Common questions.

Is AI training mandatory for educational institutions?

Yes, if the school uses AI systems in its work. Since 2 February 2025 it must take AI literacy measures for its staff and for anyone using AI on its behalf (Article 4 of Regulation (EU) 2024/1689). Since the Omnibus it does not have to guarantee a specific level, but it must take measures and be able to show them.

Is a certificate required?

No. The European Commission states that no certificate is required and that an internal record of training and other measures is enough.

How many hours of training are needed?

No rule sets hours or a syllabus. Measures must be proportionate to what each person already knows, how they use AI and who it is used on.

Is there a fine for not training staff?

Article 4 has no fine of its own in the AI Act: penalties are set by each Member State. In Spain, the law that was to set them lapsed in October 2026 when Parliament was dissolved. Even so, national authorities have supervised Article 4 since August 2026, and according to the Commission a lack of training may be taken into account after an incident.

Does it count if teachers use ChatGPT or another chatbot?

If they use it for their work, yes: the school is the deployer of that use and should include it in its measures. If nobody knows which tools are in use, the first step is to inventory them.

Do students have to be trained?

Not under Article 4, which concerns staff and anyone using AI on the school's behalf. Students are among the people AI is used on, and that must be taken into account when deciding staff training. Teaching AI to students is a separate, curricular matter.

How often should it be repeated?

No interval is set. It is sensible to review it every school year and whenever a new tool or new staff arrive.

With ALEROM AI

From spreadsheet to living inventory.

The template is enough to get started. ALEROM AI School starts from the same inventory, assigns each role the training it needs as a task with an owner, keeps the evidence and reminds you of each review. It does not certify compliance: that responsibility stays with each school.

Training is coming soon too. Video tutorials by role on the school's own tools, with a certificate of completion and an automatic record as evidence.

Sources

  1. Regulation (EU) 2024/1689 on artificial intelligence (EUR-Lex), Arts. 3, 4 and 26
  2. Regulation (EU) 2026/1744, Digital Omnibus on AI (Official Journal, 24 July 2026)
  3. European Commission: AI literacy questions and answers
  4. Regulation (EU) 2016/679, GDPR (Arts. 35 and 39)
  5. AI Act regulation and deadlines, on this site

General information, updated 25 October 2026. Not legal advice.