Training tailored to each role and level of risk, helping people use AI responsibly and giving you a record of the training.
Trusted by leading organisations
AI literacy is the knowledge and understanding people need to use AI systems responsibly and recognise the risks. Article 4 of the AI Act requires organisations to take measures to promote it. Data Science Lab provides training and workshops for managers, technical teams and employees who work with AI, tailored to their roles and the systems they use.
Consider an organisation that had everyone attend a general session on ChatGPT in early 2025. A year later, a chatbot answers customer questions, Copilot is used to draft communications, and a policy team uses a risk score built by the data team. The session covered none of those systems, and no one knows who attended which training.
The risks lie in the differences. If you rely on a summary produced by a language model, you need to know that it can invent facts. If you use a risk score to select case files, you need to know how the score is calculated and when to depart from it. These are different questions, and they call for different training.
The rule itself has changed too. Since 27 July 2026, Article 4 has required measures that support the development of AI literacy, without requiring a specific level for each person. The obligation remains in place, the European Commission emphasises, and supervision and enforcement apply from 2 August 2026.
Article 4 of the AI Act has applied since 2 February 2025. The text was replaced on 27 July 2026: you must take measures that support the development of AI literacy, but you do not have to ensure a specific level for each person.
The obligation applies to your staff and to others who use AI systems on your behalf, such as contractors or service providers. The European Commission says it also covers employees who use ChatGPT to write or translate texts.
The Commission identifies four steps: consider what AI is and how it works, your organisation’s role, the risks your systems pose, and what people already know in the context in which they use the system. According to the Commission, this also includes legal and ethical aspects.
An HR employee using a profiling assessment needs more knowledge than a colleague who occasionally has a text rewritten. According to the Commission, different levels of training or ways of learning may be appropriate.
You do not have to measure each person’s level of knowledge. You can, however, keep an internal record of the training and other initiatives you provide. The AP also recommends documenting your approach and progress.
If you use a high-risk system, the people responsible for human oversight must have the necessary competence, training and authority. For systems listed in Annex III of the regulation, this obligation applies from 2 December 2027.
We’ll look at your AI systems and roles together and propose a suitable approach.
We start with the AI systems you use or build and the people who work with them. If you already have an inventory for the AI Act, we build on it.
Through conversations or a short questionnaire, we establish what each group knows about AI, the systems they use and the risks involved.
A leader needs to make decisions and set direction, a developer to test and document, and a user to recognise errors and know when to intervene.
We provide AI training for management, technical training for teams and the Responsible AI workshop, using examples from your own processes.
You record who attended which training, what agreements apply and why you chose this approach. That record shows what you have done.
When you introduce a new system, change an objective or receive new recommendations from the European AI Board, you adjust the programme. We can also train new groups.
The Dutch Data Protection Authority recommends taking a strategic approach to AI literacy with a multi-year action plan: formalise it at leadership level, set a budget, appoint someone responsible and report on progress regularly. The AP published this guidance before the wording changed in July 2026, but the approach remains useful. New systems and new colleagues can quickly make a one-off course outdated.
A programme works best when it connects to what you already do: onboarding, buying new AI tools and the agreements in your AI governance. At Albert Schweitzer Ziekenhuis, we spoke with key people in healthcare, policy and IT to assess data maturity. This produced a baseline assessment and a path for improvement. The same principle applies to AI literacy: first, you need to know where you stand.
In a use case workshop, teams also decide together where AI is useful in their work and where it is not. For municipalities, we offer a Responsible AI workshop for municipalities, where leaders, policymakers and operational teams agree on how to proceed. You can find our full offering under training and adoption. Your legal team or compliance department should assess whether your approach meets the legal requirements.
Any organisation can launch an AI pilot. The challenge is what comes next: making AI part of the organisation. That’s where Data Science Lab makes a difference.
Read more about us
10+
8.7
200+
The AI Act defines it as the skills, knowledge and understanding that enable providers, deployers and affected people to use AI systems in an informed way and be aware of their opportunities, risks and potential harm. The Autoriteit Persoonsgegevens stresses that it is not just about technology, but also about social, ethical and practical aspects, such as interpreting a system’s output correctly.
Yes. Article 4 of the AI Act has applied since 2 February 2025 to organisations that provide or use AI systems. Since 27 July 2026, the obligation has been to take measures to support the development of AI literacy; you do not have to guarantee a particular level for each person. The Commission and Member States must help organisations with this, particularly SMEs.
Oversight rests with national market surveillance authorities, not the European AI Office. In the Netherlands, the Autoriteit Persoonsgegevens is the intended regulator, but had not yet been designated as of October 2026. That designation will follow through the act implementing the AI Act, which is not yet in force. According to the Commission, sanctions must be proportionate, although enforcement is more likely if an incident occurs because people were not properly trained.
No. According to the European Commission, a certificate is not required, and you do not need to measure employees’ knowledge levels. You can keep an internal record of the training and other initiatives you have provided. The Autoriteit Persoonsgegevens also recommends documenting your approach and progress so you can show what you are doing.
Often not. The Commission says that relying solely on the instructions, or asking employees to read them, is ineffective in many cases. Training and guidance should suit the audience’s knowledge and the context in which the system is used. According to the Commission, the instructions are insufficient for human oversight of high-risk systems.
AI literacy is one article in a broader regulation. You also need to consider how your systems are classified by risk, transparency obligations and, for high-risk systems, requirements for human oversight and documentation. On our AI Act page, you can read how we help organisations assess these requirements as a whole and establish governance.
If your organisation only uses a few AI features in standard software and no high-risk systems, a multi-year programme is probably more than you need. A focused session on the risks of those tools, clear rules for their use and a brief record of what you have done are a reasonable starting point. Expand your approach when you introduce new systems or more sensitive applications.
E-mail info@datasciencelab.nl
Phone +31 (0)20 3636 724
Together, we’ll look at your AI systems, the roles involved and what’s already been done. Then we’ll suggest which training suits each group.
Message sent
Thanks for your message. We’ll be in touch as soon as possible.
