AI literacy under Article 4 of the EU AI Act: what companies must do

Update of 15 October 2026. When I wrote this text in May, the Digital Omnibus had not yet been adopted. It is now binding law: Regulation (EU) 2026/1744 of 8 July 2026, published in the Official Journal on 24 July 2026, in force since 27 July 2026. It rewrote Article 4, the subject of this article. The answer, the section on what Article 4 requires and the questions now give the wording in force since 27 July 2026; the original wording is kept alongside for comparison. I also corrected one date: the Council adopted the Omnibus on 29 June 2026, not on 13 May.

Among all the obligations of the EU AI Act, one applies broadly and early, and is comparatively cheap to meet: AI literacy under Article 4. It has applied since 2 February 2025, to providers and deployers of AI systems alike.

What does Article 4 require?

Since 27 July 2026, Article 4 reads as amended by the Digital Omnibus: providers and deployers of AI systems take measures to support the development of AI literacy among their staff and other persons dealing with the operation and use of AI systems on their behalf. They do not have to guarantee a specific level of AI literacy for any individual. The measures take into account the technical knowledge, experience, education and training of the people involved, the context the AI is used in, and the people the AI is used on. The original wording still required measures to ensure, to their best extent, a sufficient level of AI literacy.

AI literacy, as the Regulation defines it, is the skills, knowledge and understanding that allow people to deploy AI in an informed way and to be aware of its opportunities, risks and possible harm.

Who is affected?

Practically every company that uses AI. A “deployer” is anyone who uses an AI system under their own authority in a professional context, from the marketing team using a text generator to the workshop using an AI-based inspection tool. Purely personal, non-professional use is excluded.

What good measures look like

• Basic training: how AI works, what it can and cannot do, where its limits and risks are.
• Role-specific depth: people who make decisions with AI need more than occasional users.
• Practical rules: what may be entered into which tool, how to handle outputs, when a human has to check.
• Documentation: keep a record of what training happened and who took part.

What the Digital Omnibus changes

Even the original text framed Art. 4 as a proportionate best-efforts duty (“to their best extent”) and not as a rigid training guarantee. The Digital Omnibus, Regulation (EU) 2026/1744, in force since 27 July 2026, rewrote it. Instead of ensuring a sufficient level, providers and deployers now take measures to support the development of AI literacy, and the text says expressly that no specific level has to be guaranteed. A new second paragraph obliges the Commission and the Member States to support companies, especially SMEs, and the Commission publishes practical examples. The duty is not abolished, and it has applied since 2 February 2025. Anyone who trains sensibly and documents it is on the safe side.

AI literacy is the cheapest obligation in the AI Act, and one of the most effective. Teams that understand AI make fewer expensive mistakes.