AI literacy (Article 4 of the AI Act): what it requires today and how to comply without a big programme
Article 4 is the AI Act obligation that affects almost every company: using AI at work is enough. The Digital Omnibus softened it, but did not remove it. Here is what it requires today and how to comply with little effort.
What Article 4 says today
The original AI Act text, applicable since 2 February 2025, required providers and deployers to ensure a sufficient level of AI literacy among their staff. Regulation (EU) 2026/1744 (the Digital Omnibus), in force since 27 July 2026, replaced that article.
The obligation is now to take measures to support the AI literacy of staff and of people who use AI systems on the company’s behalf. The text itself makes clear that it does not require guaranteeing a specific level for each person. The Commission and Member States must also support companies, especially SMEs, with practical examples.
Put simply: you are no longer judged on the result (“does your team know enough?”) but on the effort you can show (“what have you done to help them learn?”).
Who it applies to
Any organisation that develops or uses AI systems in its professional activity, not just high-risk AI. If your team uses an assistant such as ChatGPT or Copilot for work, you are a “deployer” and Article 4 applies to you.
If you also use high-risk AI, there is a more specific requirement: Article 26 requires human oversight to be assigned to people with the necessary competence, training and authority.
What the training should cover
The Commission published a set of questions and answers on Article 4. As a minimum, training should take into account:
- A common foundation: what AI is, how it works and which systems the company uses, with their opportunities and risks.
- The company’s role: whether it provides the system or only uses it.
- The risk of each use: what each person needs to know to recognise the risks of their tool, such as generative AI hallucinations.
- Each person’s profile: their technical experience, education and the context in which they use AI.
According to the same Q&A, no external certificate is needed, and pointing staff to the tool’s instructions is usually not enough.
How to comply in five steps
- Take inventory of the AI tools actually in use, including shadow AI.
- Group people by profile: who uses the tool, who checks its output and who decides to buy it.
- Train with your own cases: short 45–60 minute sessions with examples from your work, not general theory.
- Put the rules in writing in an AI use policy: which tools, with which data and what is prohibited.
- Document it: dated, versioned materials, attendee lists and session dates. Repeat when the tools change.
Why start with data
AI literacy rests on data literacy. Someone who can’t tell whether a figure is reliable will struggle to tell whether an AI answer is. That is why AI training should include the basics of data quality: where information comes from, what can be wrong with it and how to check it.
Common mistakes
- One generic video for everyone, and the topic is closed.
- Training without rules: explaining the tool without saying what data can go into it.
- Keeping no evidence: if it isn’t documented, it’s hard to show it was done.
Frequently asked questions
Are there fines for breaching Article 4?
The AI Act sets no specific fine for Article 4. Penalties depend on each Member State’s rules, and since August 2026 national authorities can supervise compliance.
Do I need a training certificate?
No. According to the Commission, no external certification is needed. You just need to be able to show the measures you have taken.
Does it apply if we only use ChatGPT?
Yes. Using an AI system at work makes you a deployer, and Article 4 applies to everyone, not only to high-risk AI.
Informational content, not legal advice.
Which AI Act obligations apply to you?
Free assessment with your priority AI Act gaps, plus the free AI use policy generator to set the rules.