AI literacy: what it is, what Article 4 has required since the omnibus and what to put in place now

Is AI literacy still mandatory after the omnibus?

Yes. The omnibus rewrote Article 4; it did not delete it. Providers and deployers of AI systems are still obliged to take measures. In its questions and answers on AI literacy, the European Commission writes that AI literacy remains an obligation for providers and deployers, but that a specific or 'sufficient' level is no longer mandated.

In the Act, a deployer is the organisation that uses an AI system under its own authority. If your team works with ChatGPT or Copilot, a planning tool with AI or image recognition in quality control, you fall under that definition. So you do not have to build the AI yourself.

What exactly does Article 4 ask for now?

Paragraph 1 of the new Article 4 reads: 'Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.'

In practice, that means three things:

• You take measures. If you do nothing, you do not comply. The obligation has become lighter, not optional.
• The measures fit the people and the use. A planner who assesses AI-generated routes every day needs different knowledge from a colleague who occasionally has a text summarised. People who work with the systems on your behalf are also covered by the text. The Commission gives a contractor, a service provider or a client as examples.
• You do not guarantee an outcome. You do not have to prove that every employee reaches a specific level. You must, however, be able to show what you have done.

What has changed compared with the old text?

The old text asked for measures 'to ensure, to their best extent, a sufficient level of AI literacy'. The new text asks for measures 'to support the development of AI literacy'. Two paragraphs have also been added:

• Paragraph 2: the Commission and the Member States support and facilitate the efforts of providers and deployers, in particular SMEs. To that end, the Commission publishes practical examples of compliance on the AI Act's single information platform.
• Paragraph 3: taking European competence frameworks into account, the AI Board adopts recommendations to support the Commission and the Member States in promoting AI literacy, including by setting out common objectives.

Why was Article 4 amended?

The EU legislator explains this in recital 8 of Regulation (EU) 2026/1744. According to that recital, experience shared by stakeholders shows that stringent obligations to ensure a sufficient level are not suitable for all types of providers and deployers. Such obligations, the recital says, also create an additional compliance burden, particularly for smaller enterprises. The same recital states that AI literacy should be a strategic priority, regardless of obligations and potential sanctions.

That last point is the key one for a management team. The legislator has removed the pressure, not the importance. A team that does not understand what an AI tool can and cannot do makes the same mistakes, with or without a statutory level.

What is not required?

The European Commission is clear about this in its questions and answers:

• No certificate. According to the Commission, an internal record of training and other guiding initiatives is sufficient.
• No knowledge test. Article 4 contains no obligation to measure employees' AI knowledge.
• No separate role or committee. Asked whether an AI officer or an AI governance board is needed, the Commission answers that no specific governance structure is mandated.
• No prescribed course or number of hours. According to the Commission, there is no one-size-fits-all approach, and no strict requirements or mandatory training are imposed. Nor, according to the Commission, does replicating examples from the European repository of AI literacy practices automatically give rise to a presumption of compliance: the measures must fit your own situation.

Can you be fined over Article 4?

Article 99(4) lists the provisions for which fines of up to €15 million or, if higher, 3 percent of worldwide annual turnover apply. Article 4 is not on that list, even after the omnibus. Stories about a €15 million fine for a missed training session are therefore not based on the text of the Act.

That does not mean nothing can happen. Under Article 99(1), as amended by the omnibus, the Member States lay down rules on penalties and other enforcement measures for infringements of the Regulation. These may also include warnings and non-monetary measures, and the Member States take the interests of SMEs into account. The Commission writes that supervision of Article 4 lies with the national market surveillance authorities, that the rules on supervision and enforcement apply from 3 August 2026, that penalties must be proportionate, and that a penalty is more likely if an incident can be shown to have been caused by inadequate training or instruction.

In the Netherlands, the Uitvoeringswet AI-verordening (the Dutch act implementing the AI Act), which is to give the supervisory authorities their powers, is not yet in force. The bill was open for public consultation online until 1 June 2026, and in late August 2026 the Autoriteit Persoonsgegevens (Dutch Data Protection Authority, AP) still referred to it as a draft bill. Under the bill, the Autoriteit Persoonsgegevens and the RDI (Dutch Authority for Digital Infrastructure) are given a coordinating role. What that means in practice is explained in our article on supervision since 2 August. The short version: the obligation applies, but the national enforcement toolkit is not yet complete.

Do more rules apply to high-risk AI?

Yes. Separately from Article 4, deployers of high-risk AI systems are also subject to Article 26(2): they assign human oversight to people 'who have the necessary competence, training and authority, as well as the necessary support'. The omnibus did not amend that article, and the Commission stresses that this training obligation remains in place.

The date has moved, however. The high-risk requirements apply to Annex III systems, including AI in recruitment and selection, from 2 December 2027, and to AI in Annex I products from 2 August 2028. You can read more about that shift in our article on what the omnibus changes for SMEs.

What is it sensible to put in place now?

Being able to demonstrate what you have done remains the key, under the new text too. In practice, that comes down to four steps:

• Establish who in your organisation works with AI and in which processes, including AI built into software and machinery.
• Train those employees in line with their role: what AI can do, what it cannot do, where the risks lie and which practices are prohibited. How that can be done in one day on your own work processes, as in-company AI training at your own location, is explained on the page about the AI literacy training.
• Record who was trained, in what, when and by whom. If you work with a quality system such as HACCP, FSSC 22000 or GMP+, it makes sense to record this in the same way as your other training.
• Adopt a short AI usage policy: which tools are permitted, what may and may not go into a prompt, and when the human decides.

Where can you find free help?

The new paragraph 2 requires the Commission to publish practical examples of compliance. To do so, the Commission is building on its existing repository of AI literacy practices. In the Netherlands, the Autoriteit Persoonsgegevens published the guidance documents 'Aan de slag met AI-geletterdheid' (Getting started with AI literacy) and, on 23 October 2025, 'Verder bouwen aan AI-geletterdheid' (Building further on AI literacy), with a multi-year action plan and practical examples. In them, the AP stresses that growing use of AI also calls for growing efforts on AI literacy. That guidance predates the omnibus, so read it alongside the new text of Article 4.

Is there a subsidy for AI literacy?

Possibly for part of the work around it, but not for the course itself (for what our training costs, see our offer and prices). SLIM mkb (the Dutch subsidy scheme for learning and development in SMEs) covers 60 percent of the eligible costs of a company review that results in a training or development plan, and of guidance in introducing a learning and development method. For career or development advice, a fixed amount of €700 per completed advice track applies. The maximum is €24,999 per application, or €20,000 for agricultural businesses. Training costs are explicitly excluded from the scheme, as are costs arising from legally required tasks (Article 2.15, points (h) and (g), of the SLIM regulation). We do not know whether Uitvoering van Beleid (the agency that runs SLIM) regards a training plan intended to comply with Article 4 as such a task. So describe an application in terms of your company's development needs, and raise any doubts with Uitvoering van Beleid in advance.

For training, Uitvoering van Beleid refers to the separate SLIM-scholingssubsidie (SLIM training subsidy). That subsidy is intended for training towards jobs in sectors such as engineering, construction and energy, healthcare, ICT and the green sector, so it is not a general scheme for AI training. Both SLIM mkb application windows in 2026 have closed; the August and September window was heavily oversubscribed. Windows for 2027 are not yet listed on the Uitvoering van Beleid page.