The EU AI Act's AI Literacy Requirement: What Small Teams Actually Need to Do

Most of the EU AI Act is about high-risk systems and the companies that build large models. One provision is different: Article 4, AI literacy, which has applied since 2 February 2025 and reaches almost every organisation that uses AI tools at work, including small businesses using a chatbot to draft emails.

It is also one of the vaguest provisions in the Act, which has led to two bad reactions: ignoring it, or buying expensive certification courses. Neither is necessary.

What Article 4 Says

The text of Article 4 asks providers and deployers of AI systems to take measures on the AI literacy of their staff and others operating AI on their behalf, taking into account their technical knowledge, experience, education and training, and the context the systems are used in.

“Deployer” is the Act’s word for an organisation using an AI system in a professional capacity. If your team uses AI writing tools, transcription, or chat assistants for work, you are a deployer.

Note that the wording and the balance of responsibility have been under discussion as part of the EU’s 2025 Digital Omnibus simplification proposals, which suggested shifting more of the effort to the Commission and member states. Check the current consolidated text before relying on any summary, including this one. See the EU AI Act digital omnibus delay explained.

What “AI Literacy” Means In Practice

The Act defines it as the skills, knowledge and understanding to make an informed deployment of AI systems and to be aware of their opportunities, risks and possible harms. For a team using AI to write, that boils down to people knowing:

1. What the tools do and do not do. They predict plausible text; they do not check facts.
2. Where data goes. Which tools send text to the cloud and which run locally.
3. What not to put in. Client data, personal data, confidential material, unless the tool is approved for it.
4. How to check output. Reading for changed facts, invented details and tone.
5. When AI is not appropriate. Decisions about people, legal advice, safeguarding records.
6. Who to ask. A named person for questions and incidents.

What Proportionate Looks Like For A Small Team

There is no prescribed curriculum, certificate or exam. Regulators have indicated that measures should fit the organisation’s size and risk. For a ten-person firm using AI for emails and documents, a reasonable approach is:

• A one-page AI policy listing approved tools and what data can go where. See how to write an AI use policy for a small team.
• A 45-minute session covering the six points above, with real examples from your work.
• A record of who attended and when, plus the materials used.
• A refresh when tools or rules change, or when someone new joins.

Higher-risk uses, such as AI in hiring or customer decisions, need more depth for the people involved.

Writing The Training Invitation

Before:

Mandatory AI training next Thursday 2pm, all staff must attend, EU requirement, no exceptions.

After:

On Thursday 15 October at 14:00 we are running a 45-minute session on using AI tools safely at work. It covers which tools we use, what information can go into them, and how to check what they produce. The EU AI Act asks organisations to make sure staff understand the AI they use, and this session is how we do that. Please attend if you can; if not, let me know and we will arrange a catch-up.

The second version explains what, why and how long, and treats people as adults. See how to write a training invitation email.

A Wrivio Context for policy and training communications could say:

Rewrite this as an internal announcement for all staff. Clear, friendly and direct, complete sentences. Explain what is happening, why, and what people need to do. Keep every date, time, rule and requirement exactly as written. Do not add penalties, deadlines or obligations that are not in the original.

Press Ctrl+Shift+Space, paste your draft, and check that requirements have not become softer or harsher than you meant.

How Local Tools Simplify The Conversation

A large part of AI literacy is teaching people which tools send data where. Using a local model for confidential work removes a whole category of risk from the training: if the rewrite never leaves the laptop, there is no transfer to explain. The training can then focus on the harder skill, which is checking the output.

Common Questions

Does Article 4 apply to a small business using ChatGPT?

If your staff use AI systems in a professional capacity, you are a deployer, and Article 4 is relevant. What counts as sufficient measures scales with your size and how you use AI.

Is there a fine for not providing AI literacy training?

Article 4 is not tied to a specific fine in the way prohibited practices are, but regulators can consider it when assessing other issues, and national rules may add enforcement. Treat it as a basic expectation.

Do we need a certified AI literacy course?

No certification is required by the Act. Short, relevant internal training with a record of attendance is a reasonable approach for most small organisations.

How often should we repeat the training?

When your tools or rules change, when new staff join, and at least once a year as a refresher is a sensible baseline.

Download Wrivio for Windows to give your team a local option for confidential writing, which makes your AI policy simpler to follow and to teach.