Article 4 of the EU AI Act requires every company that provides or uses AI systems to take measures that support the AI literacy of its staff. That is the whole duty. Since 27 July 2026 the text says explicitly that no one has to guarantee a particular level. There is no prescribed course, no certificate and no new deadline, and the Act itself sets no fine for Article 4.

What Article 4 says since 27 July 2026

Regulation (EU) 2026/1744, the Digital Omnibus on AI, rewrote Article 4. Providers and deployers of AI systems take measures to support the development of AI literacy of their staff and of 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 systems are used in. The sentence that matters comes next: the obligation does not require anyone to guarantee a particular level of AI literacy for any person.

The original 2024 wording asked companies to ensure, "to their best extent", a "sufficient level" of AI literacy. That formula is gone. What remains is a duty to act, not a duty to achieve a result.

What the duty does not mean

No prescribed training. The Act names no format. Training is often the obvious route, but on-the-job guidance, an accompanied start or a clear rule for a specific process can fit as well.

No certificate. In its questions and answers on Article 4 the Commission states that no certificate is needed. Companies can keep a record of what they did. That is sensible, because it shows that you acted, but it is not a legal requirement.

No fine in the AI Act itself. Article 99 does not list Article 4. Sanctions could only come from national law. Germany's implementing law, the KI-MIG, in force since 29 July 2026, sets fines for other duties but none for Article 4. If you operate in several Member States, check the rules there.

No new deadline. The duty has applied since 2 February 2025. According to the Commission, national market surveillance authorities have supervised it since August 2026. No further date is coming.

Who it applies to

Article 4 applies to all providers and deployers of AI systems, whatever the risk level. A company whose staff use a chatbot for drafts or a translation tool is a deployer and is covered. The duty covers your own staff and other persons who operate or use AI systems on your behalf, such as contractors.

High-risk systems add more: deployers must assign human oversight to people with the necessary competence, training and authority (Article 26(2)). For Annex III systems those duties apply from 2 December 2027.

What works: behaviour in the process, not course completions

The duty is narrow; the task behind it is not. A completion record from a learning portal shows that someone saw the material. It does not show whether a claims handler recognises an AI suggestion that contradicts their experience and escalates it instead of accepting it.

What works starts from a specific process:

  • For each role, name the point where AI produces a result someone acts on, and the decision that depends on it.
  • Set a rule for the doubtful case: when an AI result is checked, when it is escalated and when it is discarded.
  • Record what changed: which measure, for which role, when, and what then shows up in the process.

For a mid-sized company that is not a large training programme. It is a list of the places where AI already shapes decisions, with one fitting measure for each. The diagnostic shows in ten minutes where those places are in your organisation.

Legal status as of 11 October 2026. Takes into account Regulation (EU) 2026/1744 and Germany's KI-MIG. This article is not legal advice.