Article 4 of the AI Act has applied since 2 February 2025. In its original form it required providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy among their staff and others operating AI systems on their behalf. Regulation (EU) 2026/1744 replaced the article in full. The duty is now to take measures that support the development of AI literacy, and the text states explicitly that it does not require any specific level of literacy to be guaranteed for any individual.
What changed, and what did not
The change is to the legal character of the obligation, not its scope. It moved from a result to be achieved to an effort to be made. The people it covers, the organisations it applies to and the factors to be taken into account are the same: technical knowledge, experience, education and training, the context in which the systems are used, and the people on whom they are used. The Commission and Member States are now also tasked with supporting organisations, particularly SMEs.
The practical consequence is that the question an auditor or regulator asks has shifted. It is no longer whether each person reached a threshold. It is whether the organisation took measures that fit the AI it actually uses and the people who use it.
What still does not count
- A single generic course for everyone, disconnected from the tools people actually use.
- A certificate on its own. None is required, and a certificate does not show that the measures fit your systems.
- Measures taken once and never revisited, while the AI tools in use keep changing.
A defensible programme
- Start from the inventory. Literacy measures should map to the AI systems in use and the roles around them.
- Segment by role. Executives need to understand risk and accountability. Product and engineering teams need to understand failure modes and evaluation. Everyday users need to know what a tool can and cannot do, and when to escalate.
- Make it tool-specific. Short, practical guidance on the systems people use beats abstract theory.
- Include the affected. Where AI is used on customers, patients or employees, staff need to understand the impact on those people.
- Keep records. Document what was provided, to whom and when. The duty is one of effort, so evidence of effort matters.
- Review on change. New tools, new uses or new risks should trigger an update.
Why it still matters
Literacy is rarely the headline in an investigation, but gaps tend to surface as aggravating factors when something else goes wrong. More importantly, people who understand the tools they use make fewer costly mistakes with them. That was the point of Article 4 before the amendment, and it still is.
