What the text requires

European Union law sets an artificial intelligence literacy obligation in Article 4 of Regulation (EU) 2024/1689, known as the “AI Act”, in its consolidated version. Its first paragraph addresses two precise audiences: providers of artificial intelligence systems, those who develop them or place them on the market under their own name, and deployers, those who use them under their own authority, outside strictly personal use.

It asks them to take measures to foster the development of AI literacy among their staff and the other people who, on their behalf, use or operate those systems, taking into account their technical knowledge, their experience, their training, the context of use and the people the system is applied to. The text itself states that this obligation does not require providers or deployers to guarantee a set level of AI literacy for every individual: it is an obligation of means, not of result. It applies whatever the risk level of the system concerned, not only to high-risk systems.

This obligation has applied since 2 February 2025. Its wording has changed since: Regulation (EU) 2026/1744 of 8 July 2026, known as the digital omnibus on artificial intelligence, rewrites Article 4 entirely. It replaces an obligation of result, ensuring a sufficient level of AI literacy, with the obligation of means described above. That text has been in force since 27 July 2026: it is the one that counts today, whatever an older page may still say about it.

What the Commission recommends, without requiring it

Article 4 also charges the European Commission and the Member States with supporting this effort, in particular for small businesses, and charges the AI Board with adopting recommendations. A recommendation is not an article of a regulation: it helps to understand the obligation, it does not extend it.

The Commission’s questions and answers, updated on 27 July 2026, make clear that no measurable threshold is required and that the effort asked for is proportionate to the role of the organisation, to the risk of the system and to the training its staff have already received. They place a contractor, a subcontractor or a client among the other people covered by the article, depending on the risk of the system. They point out that the obligation holds for any organisation, whatever its size and its sector.

The Commission also keeps a public repository of practices already followed by organisations for this effort. It states itself that, as things stand, reproducing a practice from that repository does not amount to a presumption of compliance with Article 4: a repository of practices helps to understand, it ticks no box on behalf of the organisation that has to comply.

What the text does not say

What Article 4 does not say counts as much as what it does: that is where the misunderstandings sold by suppliers take hold. Neither the text of the regulation nor the Commission’s questions and answers mention, as of the date of this page:

• A certification. The word appears in neither text.
• An official attestation. The Commission writes that a simple internal record of training and initiatives is enough, with no certificate to obtain.
• Compulsory paid training. The Commission writes that no format is imposed and that no training is compulsory.
• An approved or accredited provider. Neither text provides for a list of recognised suppliers.

A supplier selling an “Article 4 certification” or an “official approval” is selling something the text does not ask for.

What this course brings to anyone who has to comply

This course provides the material that Article 4 asks to be passed on: how an artificial intelligence system builds its answer, what becomes of what is handed to it, and when not to rely on it. Three chapters of the universal core answer that directly.

• What you give it: what becomes of a text, an image or a file handed to a system, a question faced by any deployer that puts information about its staff or its clients through one.
• Where the biases come from: why a system reproduces imbalances it did not choose, one of the risks Article 4 asks people to be able to recognise.
• When not to use it: three tests for deciding whether a situation allows anyone to rely on a machine’s answer.

This course issues no attestation, keeps no training record and claims to prove no organisation’s compliance with Article 4: it explains the material, and the proof that the effort was made remains the responsibility of the organisation charged with it. The for work path offers an order for going through this course in a professional setting.

A text that can be read without an intermediary

The obligation exists, and nothing forces anyone to discover it through an intermediary: the text of Article 4 runs to three short paragraphs, it is public, and its address is cited on this page. Anxiety that pushes someone into buying compliance in a hurry deserves a reading of the text itself first.

A module apart, not the core

This obligation holds for the European Union; other regions of the world have or will have their own rules, and they will enter this course the same way, as a module placed beside the core, never inside the core itself. See the page the official frameworks, which distinguishes what this course teaches from what a law or a competency framework asks for.

Sources

• Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence, consolidated text, Article 4 “AI literacy”, European Union, EUR-Lex, 2024, text consolidated as at 27 July 2026.
• Regulation (EU) 2026/1744 of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (the digital omnibus on artificial intelligence), European Union, EUR-Lex, 2026, published on 24 July 2026.
• AI Literacy, Questions & Answers, European Commission, AI Office, 2025, updated on 27 July 2026.