EU AI Act Article 4: planning AI literacy at work

The short answer on Article 4

Article 4 of the EU AI Act requires providers and deployers of AI systems to take measures that support the development of AI literacy among staff and other people who operate or use those systems on their behalf. The measures should consider technical knowledge, experience, education, the context of use and the people affected by the systems.

Article 4 has applied since 2 February 2025. The amended version entered into force on 27 July 2026; it does not require a guarantee that every individual reaches a specified level of AI literacy. Organisations still need to act. The wording underlines that a relevant, context-specific approach matters more than one identical course for everyone. This guide supports planning and is not legal advice.

Start with your role, systems and affected tasks

Establish whether your organisation provides AI systems, deploys them in its operations, or does both. Then create a manageable inventory of the systems actually in use and the tasks in which people work with them. A broad label such as “generative AI” is not enough. What matters is whether a tool drafts content, supports candidate screening, contributes to forecasts or generates software code.

For each use case, record the roles involved, permitted data, expected outputs, known failure modes and the person responsible for review. Separately assess whether rules for high-risk systems or other areas of law apply. Article 4 alone is not a complete compliance checklist.

Define what each role actually needs to know

The European Commission recommends considering the organisation’s role, the risks of its systems, existing knowledge and the actual context of use together. From this, you can build a learning matrix covering role, system, typical decision, required capability, learning measure and owner. This matrix is an editorial planning aid, not an official template.

• Which AI systems does the role use or develop?
• Which decisions are supported or made?
• Which errors, bias or data protection risks must the person recognise?
• Who reviews outputs, and when should use stop or be escalated?

Design measures around real work situations

A short foundation module can establish shared terminology, opportunities, limitations and internal rules. Different roles then need different practice. A marketing team might check unsupported claims and sensitive inputs, developers can test and document code suggestions, and managers can work through approvals and responsibilities using realistic cases. Use approved tools and fictional or explicitly suitable training data.

Use observable learning outcomes. Instead of “understand AI”, a goal could be: the participant identifies three common failure modes of the system, checks an output against its source and records who approves the decision. These outcomes can be demonstrated in an exercise without promising legal compliance.

Keep a clear record of learning measures

According to the European Commission FAQ, Article 4 does not require a particular certificate. Organisations can keep internal records of training and other guidance initiatives. For internal traceability, you could record the audience, systems and risks covered, learning outcomes, format, date, participation and agreed next steps. Adapt the record to your situation and internal retention rules.

A record alone does not prove that a measure worked. Include questions, practical tasks or a later review. The examples in the Commission’s practice repository do not by themselves create a presumption of compliance. They offer orientation and need to be adapted to the organisation’s own context.

How to assess an AI literacy course

Ask providers for an agenda that visibly addresses your roles, systems and use cases. Confirm prerequisites, the amount of hands-on practice, the materials available for internal follow-up and what is explicitly outside the scope. A responsible course does not promise blanket legal compliance or an automatically sufficient level of knowledge.

NextNowa confirms content, pricing and dates on request. Include team size, roles, AI systems in use, typical tasks, existing policies and desired learning outcomes. Before booking, the delivery scope, responsible instructors, availability and included services are confirmed in writing.