Blog · EU AI Act
AI literacy: what Article 4 of the EU AI Act really requires
If anyone in your company uses ChatGPT, a machine translator or software with AI built in, Article 4 of the EU AI Act applies to you. The good news: since the July 2026 Digital Omnibus it asks for less than you are being sold.
In a nutshell
Article 4 requires providers and deployers of AI systems to take measures to support the AI literacy of their staff. It has applied since 02/02/2025.
Since the Digital Omnibus (consolidated version of 27/07/2026) the text no longer asks to «ensure a sufficient level»: it expressly says it does not require a specific level for any individual.
It doesn't require a qualification, minimum hours or an approved certificate. It requires measures proportionate to actual use and being able to show they were taken.
This article is accurate as of: 11/11/2026.
What Article 4 says today
This is the text in force, in the consolidated version of Regulation (EU) 2024/1689 as of 27/07/2026 (our translation of the official Spanish version; check the official English text on EUR-Lex):
«Providers and deployers of AI systems shall take measures to support the promotion of AI literacy of their staff and 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 AI systems are to be used in […]. This obligation does not require providers or deployers to ensure a specific level of AI literacy of any particular person.»
The same Article adds two things: the Commission and the Member States will support companies (particularly SMEs) and the Commission will publish practical examples of how to comply.
What changed with the Omnibus
The original 2024 wording required measures «to ensure, to their best extent», that staff had «a sufficient level of AI literacy». Regulation (EU) 2026/1744, known as the Digital Omnibus, replaced it with «measures to support the promotion» and added the sentence saying that no specific level is required.
| 2024 text | Text in force | |
|---|---|---|
| What is required | Measures to ensure a sufficient level | Measures to support the promotion of AI literacy |
| Level of each person | «Sufficient», to the best extent | No specific level required |
| Public support | Not mentioned | The Commission and Member States provide support, especially to SMEs |
Many courses and slide decks still quote the 2024 text. If someone talks about «ensuring a sufficient level», they are using the superseded version.
Who it applies to
To providers (who develop an AI system and place it on the market under their own name) and to deployers. The Regulation defines the latter (Art. 3(4)) as anyone using an AI system under their own authority, except in a personal non-professional activity.
In practice, almost any company that uses AI in its work is a deployer: the accounting firm that summarises contracts with an assistant, the haulier whose routing software has AI built in or the shop that answers customers with a chatbot.
What it doesn't require
- It doesn't require a qualification or an official certification.
- It doesn't set minimum training hours.
- It doesn't require an approved provider: there is no register of authorised trainers.
- It doesn't require a specific level for each person: the Article says so itself.
On fines: Article 99(4) sets fines for breaching specific obligations (including those in Article 26 and the transparency obligations in Article 50) and Article 4 is not on that list. Each Member State lays down the rules on penalties and other measures (Art. 99(1)). The absence of a fine set in the Regulation doesn't mean you don't have to comply.
How to comply without pointless paperwork
An obligation of means is met with reasonable measures and a record of having taken them. This is what we suggest to companies:
- Know who uses what. A simple inventory of AI tools and who uses them. Without it, nothing can be tailored.
- Train according to use. Someone drafting emails with an assistant doesn't need the same as someone using a system that scores job applicants.
- A short use policy. Which tools are approved, which data never goes into a chat and when a person reviews the output.
- Keep a record. Who was trained, when and on what. It's the proof that the measures exist.
- Review it when the tools change. An AI inventory goes out of date within months.
For point 2 we have our ALF-01 course on GestCampus: 8 hours, 8 units and a final test, recording who took it, for how long and with what score. And if you want the full package, our EU AI Act compliance consultancy sets up the inventory, the policy and the training.
Frequently asked questions
Since when has Article 4 applied?
Since 02/02/2025, like the rest of Chapters I and II of the Regulation (Art. 113).
Do I have to give my staff a certified course?
No. Article 4 requires measures to support AI literacy, tailored to actual use. A course with a record helps to prove it, but the Regulation doesn't require any particular certificate.
My company only uses ChatGPT for emails. Does it affect me?
Yes. If it's used in the course of business, the company is a deployer and Article 4 applies. A proportionate measure can be simple: usage guidelines and a short training session.
What changed in July 2026?
The Digital Omnibus (Regulation (EU) 2026/1744) softened the text (from «ensure a sufficient level» to «support the promotion») and made clear that no specific level is required of any individual.
Any questions about this?
Write to us from the contact form and a member of the team will get back to you.
Sources
- Regulation (EU) 2024/1689, Arts. 3, 4, 99 and 113, consolidated version as of 27/07/2026 (CELEX 02024R1689-20260727): https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02024R1689-20260727
- Regulation (EU) 2024/1689, original text published in the OJEU (CELEX 32024R1689): https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689