This page is a practical summary for operators and founders. It is informational, not legal advice.
1. Direct answer
If your organisation uses AI systems in work, Article 4 matters to you. The obligation is not to buy a giant learning platform. The obligation is to take measures, to your best extent, to ensure a sufficient level of AI literacy among staff and other people using AI on your behalf.
- Who is in scope? Providers and deployers of AI systems, plus relevant staff and other people acting on their behalf.
- Since when? Article 4 has applied since 2 February 2025.
- What is the practical ask? Match training to role and use context, then keep sensible proof.
2. Who it applies to
The key legal categories are providers and deployers of AI systems.
Providers
Organisations that develop an AI system or place it on the market under their own name or trademark.
Deployers
Organisations that use an AI system in the course of their professional activities.
The deployer case is usually the most common. If you use tools like ChatGPT, Copilot, Claude, or other AI tools in work, you are usually thinking about deployer-side obligations.
3. Since when
Article 4 has applied since 2 February 2025. That is the important date to remember.
The AI Act applies in stages. Not every rule started at the same time, but Article 4 is already in force.
4. What reasonable measures look like
The law does not prescribe one exact format. It expects measures that fit the users' knowledge, experience, education, training, and the context in which AI is used.
Note that you don't need to provide long courses that takes several hours to complete. Short training sessions with key things to remember are often enough for most roles.
| Good first step | Why it fits Article 4 better |
|---|---|
| Role-based training tracks | Different jobs use AI differently, so the training should reflect real context. |
| Short assessment or attestation | Helps show that people did more than just receive a link. |
| Completion records with dates | Gives you evidence later if a customer, auditor, or regulator asks. |
| Periodic refresh | AI usage changes fast, so static one-time onboarding is often weak. |
For most SMEs, a lightweight workflow is enough: identify the roles that use AI, map the main risks, train those people with short key lessons, and keep the record.
5. What proof to keep
Article 4 does not require one specific certificate format. But if you want to show that you took reasonable measures, you should keep simple proof.
- Who was asked to complete the training
- Which training track or curriculum they completed
- When they completed it
- What the training covered
- Any assessment, attestation, or acknowledgement used
A clean PDF training record is often a practical choice because it is easy to review and easy to share internally.
For more on that, see what proof to keep for AI literacy.
6. Common mistakes
- Assuming Article 4 only applies to enterprise AI projects
- Using one generic awareness video for every role
- Ignoring contractors and external contributors who use AI on your behalf
- Keeping no proof beyond a vague statement that training happened
- Quoting the wrong fine tier for Article 4
Article 4 is a legal obligation, but it does not have its own specific EU-wide fine tier in the text. Safe framing is that AI literacy is mandatory under the EU AI Act, while penalties are implemented and enforced through national systems.