EU AI Act

Do we need to do something about AI literacy?

Short answer: yes, if your people work with AI. Since 2 February 2025, the AI Act has required you to take measures so that they understand the systems they use. What that means in practice — and what it does not mean — is set out below.

What Article 4 says

The AI Act requires providers and deployers of AI systems to take measures that support the AI literacy of their staff and of others who operate those systems on their behalf, taking into account their knowledge and experience and the context in which the systems are used. Since the 2026 amendment (the Digital Omnibus), you no longer have to guarantee a specific or ‘sufficient’ level of AI literacy: you take measures that support it.

It applies more widely than you might think

Not only to those who build AI. If you use ChatGPT, Copilot or Claude in everyday work, you are a deployer (the AI Act’s term for an organisation that uses an AI system under its own authority) and it applies to you — whether or not the use is high-risk.

It is an open standard

The law sets no exam, no recognised certificate and no minimum number of hours. Which measures fit depends on the role and the risk. That leaves room, but it does call for a choice you can explain.

It already applies

Article 4 has applied since 2 February 2025, and national supervisory authorities have overseen it since 2 August 2026. The question is not whether it applies, but what you have put in place.

A sense of proportion

What the law does not ask of you

Large figures get quoted around this subject, and they deserve some nuance. The heavy fines in the AI Act belong to prohibited practices and to the obligations for high-risk systems — not directly to Article 4.

So you do not have to put everyone through an exam, and you do not need a quality mark on the wall. What does count: that you have made a deliberate assessment for each group of employees, that you have acted on it, and that you can show it when asked. The real risk is not a fine on day one, but having nothing to show after an incident.

Literacy is not a course you tick off. It is knowing when not to trust what a model tells you.

How to approach it

Three steps. You can take the first one yourself; we usually do the other two together.

Step 1

Find out who works with AI

Usually more people than you expect, including those using it outside official channels. Group them by what they do with it and what goes wrong if it fails. That distinction shapes everything else.

Step 2

Decide what each group needs

A team that drafts text with AI needs different knowledge from a team that uses AI in decisions about people. We turn that into a programme per group, on the platform you already use.

Step 3

Make it demonstrable

Record who completed what, and when. Our certification does this automatically, with a verifiable certificate per participant and an overview for the organisation.

Where we come in

From obligation to something that pays off

Most organisations start because they have to. They keep going because it works: people who understand what a model can and cannot do get more out of it and use it more carefully. The two go together.

Our training programmes follow the platform you already use — Microsoft Copilot, OpenAI or Claude — and end in a certification that documents what people know. Not a general awareness session, but working with the tools already on people’s desks.

Example of a Westcube AI literacy certificate

Frequently asked questions

Does the AI literacy obligation apply to us?

If your organisation uses or provides AI systems, Article 4 has applied since 2 February 2025. You do not have to build the technology yourself: using ChatGPT or Copilot in daily work makes you a deployer. The obligation covers all AI systems, not only high-risk ones.

Is there a mandatory exam or certificate?

No. The AI Act does not prescribe an exam, certificate or accredited course. Article 4 is an open standard: you take measures that fit the knowledge, experience and context of your people. A certificate is not required, but it is a practical way to show what you have done.

What level do our people need to reach?

There is no set level. Since the 2026 amendment (the Digital Omnibus), you no longer have to guarantee a specific or ‘sufficient’ level of AI literacy. You do have to take measures that fit who uses AI, for what, and with what consequences. A marketer who drafts copy with AI needs different knowledge from someone who uses AI in recruitment.

What happens if we do nothing?

The heavy fines in the AI Act target prohibited practices and high-risk obligations, not Article 4 directly. National supervisory authorities have overseen Article 4 since 2 August 2026, though. The real risk is practical: after an incident, or when a regulator asks, you need to show that you equipped your people. Without a record, that is difficult.

How long does it take to put this in place?

For most organisations, a first round takes a few weeks: identify who works with AI, decide per group what is needed, and set up a programme. Recording it properly takes little extra time if you set it up that way from the start.

This page explains how we read the rules and is not legal advice. For an assessment of your specific situation, please consult your legal adviser.

Request a proposal

AI literacy for your organisation

Tell us briefly where your team stands and we will be in touch within one working day. You get a tailored programme, not a standard package.

From €100 per participant per session, excluding VAT, for groups of 8 or more. The final price is set out in the proposal.

No obligation. We only use your details to respond to this request.