In July, the managing director of an engineering firm receives a quote: AI Act training, two hours per employee, certificate included for the file. Below the email, the privacy officer asks whether this is now mandatory. The honest answer: the obligation exists, the course in that form does not.
Article 4 of the AI Act has applied since 2 February 2025. It requires providers and deployers, in plain terms whoever builds an AI system and whoever uses one under their own responsibility, to take measures to ensure a sufficient level of AI literacy among their staff and other people who use AI systems on their behalf. What counts as sufficient depends on knowledge, experience, education and the context of use. There is no course in it, no certificate and no number of hours.
What does Article 4 actually say?
The text is short. Providers and deployers take measures to ensure, to their best extent, a sufficient level of AI literacy among their staff and other persons dealing with the operation and use of AI systems on their behalf. In doing so they take into account those people's technical knowledge, experience, education and training, the context in which the systems are to be used, and the persons on whom the systems are to be used. That is how it reads in Regulation (EU) 2024/1689.
The same regulation defines AI literacy as the skills, knowledge and understanding needed to make an informed deployment of AI systems and to be aware of the opportunities, the risks and the possible harm. So it is not about being able to code. It is about knowing what you are using, what it cannot do, and when not to take an output at face value.
Two words deserve attention. A "deployer" is not only a hospital running a triage system; it is also an office where people use a chatbot to draft a proposal. And "on their behalf" reaches contractors and suppliers who work with your AI systems, not only the people on payroll.
The Digital Omnibus of July 2026 reworded the provision into an obligation of effort: take measures that support the development of AI literacy. What that does and does not change is set out in the AI Act timeline after the omnibus. For the question of what to do, it makes little difference: the obligation is there, and it is still contextual.
Why is there no course in it?
Because the same provision has to work for a manufacturer of medical software and for an accounting office with eight people. One standard for both would be too light for one and too heavy for the other. The legislator chose an open norm: the organisation decides what is sufficient, and must be able to explain that choice.
The European Commission confirmed this in spring 2025 in its questions and answers on AI literacy. A certificate is not required, nor is a formal test of staff knowledge, and no fixed training package is coming from Brussels. An organisation keeps its own record of the measures it took. We could not open that page ourselves while writing this piece and rely on several legal summaries of it; read the text before you build policy on it.
The Commission also maintains a collection of practices, a "living repository" of what organisations have set up themselves: from formal courses to team discussions about what does and does not belong in a prompt. Useful to borrow from, with one caveat the Commission adds itself: adopting an example from the list does not by itself mean you comply with Article 4.
What does sufficient mean per role?
The question is not "how many hours of training" but "what does someone in this role need to understand to use AI responsibly". That differs, and those differences are exactly what you write down.
| Role | What the person needs to be able to do | What you record |
|---|---|---|
| Uses a chatbot for text work | Knows which data does not belong in it, which tools are allowed, and that an answer needs checking | The explanation this group received, and the decision per tool |
| Uses AI in decisions about people (recruitment, customer acceptance, care) | Understands how the system reaches an output, what it misses, and when to deviate | The instruction per system and who was trained on it |
| Buys or assesses AI tools (IT, procurement, privacy) | Can judge the vendor, its training policy, its country and the risk | The questionnaire and the register of assessed tools |
| Sets the policy (management, board) | Knows the statuses per tool, the line between allowed and not, and its own responsibility | The policy, the decisions and who took them |
For most organisations the weight sits in the first row. That is where the largest group is, and where the explanation is simplest: this is allowed, this is not, and this is where you go when in doubt. How the AI Act and the GDPR fit together here is covered in what the AI Act and the GDPR require of businesses.
Who supervises it?
Article 4 has applied since February 2025, but supervision came later. Since 2 August 2026 the national authorities have had the power, together with most of the regulation. Article 4 has no penalty provision of its own in the regulation; which sanction applies is for member states to set in national law.
In the Netherlands that happens in the AI Act Implementation Act (Uitvoeringswet AI-verordening). The bill went out for consultation in April 2026 and divides supervision among existing authorities, with a coordinating role for the Dutch Data Protection Authority (AP) and the Dutch Authority for Digital Infrastructure (RDI). Who exactly supervises Article 4 in the Netherlands, and with which powers, is therefore not yet final. Write that down in your own policy as well: it is no reason to wait, but it is a reason not to claim certainties that do not exist yet.
The question "can you show it" rarely comes from a supervisor first, by the way. It comes from a customer in a tender, from the auditor, from the works council, or after a data breach in which an employee pasted a file into a chatbot. What counts then is not whether there was a course, but whether you can show what the organisation had decided and whether that decision reached the employee.
What does this look like in an ordinary organisation?
An accounting firm with 120 staff has everyone follow an AI course in spring. Copilot is allowed, client data does not belong in free tools, and when in doubt you ask the privacy officer. Attendance 94 percent, the list goes in the folder.
An inventory in summer shows that alongside Copilot, ChatGPT, Perplexity and a meeting transcription tool are in use. No decision was ever taken on the transcription tool; two teams have been using it for client meetings for months. That is not an incident and not proof that the course failed. It shows that the course was one moment and the work carried on, with tools that did not exist that day or that nobody recognised as AI.
The firm then does two things. It takes a decision on the transcription tool and records it in the register. And it makes sure that whoever opens the tool also gets to see that decision.
Why knowledge is no guarantee of behaviour
That second step is the part of AI literacy that is missing most often. A course changes what people know. What they do on a Thursday afternoon, under a deadline, is shaped more by habit than by knowledge. Why that is, and what a course can and cannot do, is set out in awareness training makes people aware, it does not change their behaviour.
For Article 4 this means: passing on knowledge is the first half. The second half is that the organisation's decision is present at the moment someone opens an AI tool. Not as a block, but as a reminder: this is what was decided about this tool, this is the approved alternative, and sensitive data in your text gets a highlight before you send. Whoever sets that up no longer has to hope that an explanation from three months ago comes to mind by itself.
It also yields something for the record. Every notice can be counted with what the person did: opened the alternative or continued anyway. Every highlight with its outcome: masked, removed, or sent regardless. As counts, without who it was. That is the difference between "we gave a course" and "we know what has happened since".
