Approach · 1 min read
AI literacy training: what the AI Act asks of you
Since February 2025 the European AI Act requires employers to make sure staff are sufficiently AI literate. What does that mean in practice, and how do you arrange it without creating a paper tiger?
Since 2 February 2025 the European AI Act contains an obligation many small businesses haven't noticed yet: whoever uses AI at work must make sure the people working with it are sufficiently AI literate. Not as a vague ambition, but as a requirement in article 4 of the law.
The good news: this is not a paper monster. It is exactly what a sensible company wants anyway, namely a team that knows what AI can do, where it goes wrong and which information never belongs in it.
What AI literacy means
AI literacy is the combination of knowledge and skill to use AI sensibly. For an office team that comes down to four things:
- Knowing what an AI assistant can and cannot do, and why answers are sometimes convincingly wrong.
- Being able to give good instructions and check the outcome before it goes out the door.
- Knowing which data may and may never go into an AI tool, in line with GDPR.
- Understanding when a human must take the decision, and putting that in writing.
The law asks for no diploma and prescribes no number of hours. The bar is "sufficient for the context": a marketing team writing copy with AI needs different knowledge than an administration processing customer data with it.
Who it applies to
Every organisation that uses or offers AI, large or small. So also the twelve-person company where three colleagues work with Claude or ChatGPT daily and nobody ever made any agreements. That is exactly where the risk is largest: confidential documents in a chat window, answers going to customers unchecked.
The obligation sits with the employer. You don't have to turn every employee into an AI expert, but you must be able to show that the people working with it know what they are doing.
How to arrange it in practice
A workable approach in four steps:
- Take stock. Which AI tools are being used, including the unofficial ones? Who uses them for what?
- Make agreements. What may and may not go into an AI tool, who checks what, and where is that written down?
- Train the team. One good training day brings knowledge, skill and agreements up to standard in one go, using your own work examples so it sticks.
- Record it. Who was trained when, which certificate goes with it and where the agreements live. That file is your evidence.
Our trainings are built on this approach: every training starts with the agreements about data and ends with a certificate of participation and written team agreements. For teams that mainly need to learn to work safely with an AI assistant, the Claude training for beginners is the logical start; if you want to cover the whole organisation in one go, we come on site to your company.
Why you shouldn't postpone this
Not because of the fine, although that may come. The real reason is that untrained AI use already carries risks today: leaked customer data, errors in quotes and texts that aren't yours but do carry your name. The AI Act turns common sense into an obligation. Whoever handles it well is not only compliant but simply has a faster team.
Not sure where you stand? In a 15 minute intro call we'll tell you honestly whether a training is needed, and which one.
Frequently asked questions
Read more
Ready to hand it over?
We'll take thirty minutes to look at where most of your time leaks away. We usually reply within one business day.
Book an intro call