The short answer
Keep evidence of which AI systems people use, the groups and risks involved, what support they received and how you review whether it helps. A knowledge assessment can inform the baseline and learning plan, but no single test or certificate establishes Article 4 compliance.
Start with the current Article 4 wording
The consolidated EU AI Act text requires providers and deployers to take measures to support the development of AI literacy for staff and others operating or using AI on their behalf. It asks organisations to take account of people's technical knowledge, experience, education and training, the context in which systems are used and the people affected by them. It does not prescribe an individual test score.
The European Commission's AI literacy Q&A says there is no required certificate and that organisations can keep an internal record of training and other guiding initiatives. It also says the approach depends on the systems, people and context. Check the official text and guidance when reviewing your programme, as implementation details can change.
Build a practical evidence trail
First record the systems in use, who uses them, what decisions they affect and which people might be affected. Then record the baseline you used to choose support. This might include discussions, a survey about access and confidence, or a consistent knowledge measure such as Sova-delivered Core-6. Each method captures a different part of the picture.
Next retain the training, guidance and practice opportunities you provided, including their audience and rationale. Record the instructions for approved use, data boundaries and human oversight. Finally, review whether the support still fits the systems and tasks as they change. A record of attendance alone cannot show that a person understands a risky output or has authority to challenge it.
Core-6 measures foundational AI knowledge through one generic assessment across roles. Its six-area results can help target learning and show changes in knowledge. Task practice, manager support and governance are separate employer activities; neither the assessment nor the optional A/A/H and PRIME worksheet certifies legal compliance.
Use a proportionate review rather than a certificate claim
A useful programme record explains why measures were chosen and what happened after the results. If a group handles sensitive recruitment data, document its particular data rules and the support it receives. If another group only drafts low-risk internal text, the relevant examples and follow-up may differ. Avoid presenting a single score as proof that every person or workflow is ready.
Review the latest official text and Commission guidance when you change the programme. Record which version informed your choices so that the evidence trail reflects the rules and the tools in use at the time.
Put it into practice
A simple way to start
- ContextRecord systems, users, tasks, risk and people affected.
- BaselineCapture the knowledge or other evidence used to choose an intervention, with its limits.
- MeasuresKeep a record of learning, guidance, practice and human oversight arrangements.
- ReviewRevisit the plan when tools, roles or the evidence change.
Sources and further context
- Consolidated EU AI Act, Article 4 Current legal text and its context
- European Commission AI literacy Q&A Commission guidance on programme content, documentation and certificates