Article 4

The AI literacy obligation under Art. 4 EU AI Act

Yes, to measures: since 2 February 2025, companies that are providers or deployers of AI systems must take measures concerning the AI literacy of their staff. Art. 4 does not prescribe a specific form of training.

Since 27 July 2026, Art. 4 requires providers and deployers to take measures to support the development of AI literacy among their staff. They do not have to guarantee any particular level of competence. Until 26 July 2026, the obligation required them, to their best extent, to ensure a sufficient level of AI literacy.

Is there an AI training obligation?

Art. 4 requires measures to support AI literacy, but no particular form of training and no guaranteed level of competence. The wording of paragraph 1 as currently in force:

“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.”

Source: Art. 4(1) Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, EUR-Lex

Article 4 of the EU AI Act therefore sets out criteria, but no catalogue of learning content: account must be taken of the technical knowledge, experience, education and training of the persons concerned, the context in which the AI system is used and the groups of people it is used on. People often speak of an “AI training obligation”. Art. 4 does not expressly prescribe training; but training is a simple measure that reaches everyone involved.

Important: the AI literacy obligation applies to virtually every company that uses AI professionally, regardless of sector, company size or the risk class of the system.

What changed on 27 July 2026?

Regulation (EU) 2026/1744 (Digital Omnibus) entered into force on 27 July 2026 and reworded Article 4. Until then, providers and deployers were to take measures to ensure, to their best extent, a sufficient level of AI literacy among their staff. Since the reworded version, they take measures to support the development of AI literacy, and it is expressly stated that no particular level has to be guaranteed. In addition, the Commission and the Member States are to support companies, in particular SMEs, in implementing the obligation. The obligation remains, but its content has been reworded: it has applied since 2 February 2025, and since 27 July 2026 in the new version.

When did the obligation take effect?

Article 4 has applied since 2 February 2025, the first milestone of the EU AI Act, on which the ban on certain AI practices and the literacy obligation both became directly applicable. Since 27 July 2026 it has applied in the new version. Companies that have taken no action to date should start promptly. All further dates — including the obligations for high-risk AI from 2 December 2027 (Annex III) and 2 August 2028 (Annex I) — are set out on the deadlines page.

Who does it apply to?

Article 4 applies to providers and deployers of AI systems, that is, to the two main groups distinguished by the EU AI Act:

Providers

Anyone who develops an AI system, or has one developed, and places it on the market or puts it into service under their own name or trademark (Art. 3(3)).

Deployers

Anyone who uses an AI system under their own authority, except for purely personal, non-professional use (Art. 3(4)). As a rule, that is the company, not the individual user.

All company sizes

Whether a large corporation, an SME or a start-up: Art. 4 sets no minimum size. Exceptions to the scope are set out in Art. 2, for example for national security, research and development only, or purely personal use.

Every risk class

Even using simple AI tools — for example for writing assistance or image editing — triggers the obligation to take measures to promote AI literacy.

Anyone who only uses AI and does not develop it is also covered by Art. 4 as a deployer (Art. 3(4), Art. 4). The deployer is the company that uses AI systems under its own authority; the measures are aimed at the people who work with them on its behalf. More about the roles and the risk-based approach of the Regulation is explained on the EU AI Act overview page.

What does AI literacy mean in practice?

Art. 3(56) defines AI literacy as “skills, knowledge and understanding that allow providers, deployers and affected persons, taking into account their respective rights and obligations in the context of this Regulation, to make an informed deployment of AI systems, as well as to gain awareness about the opportunities and risks of AI and possible harm it can cause”. In practice, this means employees — appropriate to their role — should understand:

  • how AI systems work in principle, where their limits lie and why they can produce errors;
  • what opportunities and risks are associated with the specific use, including biases, so-called "hallucinations" and data protection issues;
  • how AI outputs can be critically assessed, questioned and corrected where necessary;
  • which legal and ethical frameworks — in particular the EU AI Act — apply to their own working environment.

From practice: literacy is not a one-off state, and new AI tools require new assessments. It makes sense to repeat and update training on a regular basis; Art. 4 sets no fixed deadline for this.

How do companies fulfil the obligation?

Companies fulfil the obligation with measures that take into account the knowledge, experience, education and context of use of the people involved; Art. 4 does not prescribe a specific format. Article 4 does not expressly mention documentation, but it is the simplest way to demonstrate the measures taken later on. A simple measure is a documented training course with a knowledge assessment and an individual certificate. The AI-Konform training covers the relevant subject areas in four lessons: Introduction to the EU AI Act, Risk classes, Obligations & AI literacy, and AI fundamentals. The format is deliberately compact at 30 to 60 minutes per person, so that even large teams can complete the training without lengthy downtime. After the final quiz, participants receive a personal certificate; the company automatically receives an audit report (CSV/PDF, also for Vanta and ISO 27001 audits) and a Compliance Score as evidence of the measures taken, for auditors and authorities.

  • Record AI use in the company and identify the relevant roles.
  • Train employees to suit their role and context of use: a basic course for everyone, plus an introduction to the specific systems in use, each in their preferred language.
  • Archive participation, test results and certificates in a verifiable manner.

What is behind the term “AI driving licence” is explained in our guide AI driving licence.

Proof of AI literacy: what should you document?

Art. 4 does not require formal proof of AI literacy. If an authority asks, however, you should be able to show which measures you have taken. A lean record covers:

  • which AI systems are in use and which roles work with them,
  • which measures you have taken, such as training, briefings on specific tools or an AI policy, each with date and content,
  • who took part and with what result, for example in a quiz,
  • when you refresh the measures, for example when new AI tools are added,
  • who in the company is responsible.

The European Commission publishes practical examples of how companies meet the obligation on its central information platform (Art. 4(2)).

What are the consequences of non-compliance?

For Article 4, the EU AI Act contains no fine provision of its own: Art. 99(3) to (5) do not mention Art. 4. Those fine tiers (up to €35 million or 7 % for prohibited practices under Article 5, up to €15 million or 3 % for certain other obligations) concern other infringements. The Member States lay down the sanctions for infringements of Article 4 (Art. 99(1)); these may be fines, warnings or non-monetary measures. Indirectly, AI literacy can play a role in other obligations, for example under Art. 26(2): human oversight of high-risk AI may only be assigned to persons with the necessary competence, training and authority (from December 2027 and August 2028 respectively). Regardless of sanctions, promoting AI literacy remains a strategic task. All fine tiers and the corresponding offences are explained in detail on the fines page.

Frequently asked questions about the AI literacy obligation

Is there an AI training obligation under the AI Act?

Not expressly. Art. 4 requires measures to promote AI literacy, but it prescribes no format, neither training nor a certificate. A documented training course is a simple measure, because it reaches everyone involved and records participation and result. How this works with AI-Konform is shown on the EU AI Act training page.

Is there a mandatory AI training?

No. The EU AI Act does not prescribe mandatory training with fixed content, scope or qualification. Art. 4 requires measures that take into account the knowledge, experience, education and training of the persons involved and the context of use; which measures these are is for the company to decide.

Do employees need to become AI experts?

No. Article 4 requires measures to promote AI literacy that take into account the role, prior knowledge and the AI system in use. No particular level has to be guaranteed. A clerk who occasionally uses an AI writing assistant does not need in-depth technical knowledge. What matters is that they know the tool's limits, can assess outputs critically and are aware of where the legal boundaries lie. A compact training course with a knowledge assessment and an individual certificate is a simple, well-documentable measure under Article 4.

Does the obligation apply even if we only use AI occasionally?

Yes. Art. 4 sets no minimum frequency of use. If your company uses an AI system under its own authority, it is a deployer (Art. 3(4)) and takes measures for the people who work with it, even if they only do so occasionally. Further answers to common questions about the EU AI Act, deadlines and training can be found on our FAQ page.

Do I need proof of AI literacy?

No. Art. 4 requires no formal proof, no certificate and no official certification. If you want to be able to demonstrate AI literacy under Art. 4, document the measures taken: who took part in what, and when. This is exactly what the participation certificate per person and the audit report from AI-Konform record. More on the EU AI Act training page.

Is there an official certificate for AI literacy?

No. Art. 4 does not provide for an official certificate or mandatory certification of AI literacy. Certificates from training providers, including the AI-Konform certificate of participation, document a measure you have taken; they are not an official attestation.

Who checks AI literacy?

Each EU Member State designates national market surveillance authorities that monitor compliance with the AI Act (Art. 70). Sanctions for breaches of Art. 4 are laid down by the Member States (Art. 99(1)). If an authority asks, your documented measures show what you have done. More on who is responsible on the fines page.

Train your team now Explore the training

Note: this article is for general information only and does not constitute legal advice.