AI literacy under Article 4: what really applies since the Digital Omnibus

Written By :

Category :

Insights

Posted On :

Share This :

Since February 2025, the EU AI Act has required companies to do something for the AI literacy of their staff. Many turned this into a training obligation with a certificate. Since 27 July 2026, a new version of Article 4 applies. It is clearer and, in one respect, milder. The obligation itself remains, however, including for companies that only use ChatGPT or Copilot.

This article shows what has changed, what has stayed the same and how a small or medium-sized business can meet the obligation with reasonable effort. Transparency note: an AI assistant supported us in writing this article. Status: 17 November 2026. This article does not constitute legal advice.

Key points at a glance

  • New version since 27 July 2026: providers and deployers of AI systems take measures to support the development of AI literacy of their staff. They do not have to guarantee a specific level for each person.
  • Who is affected: every company that uses AI professionally, including for writing texts with ChatGPT or analyses with Copilot.
  • What is not required: no certificate, no standard training, no AI officer, no individual proof of competence.
  • What is recommended: assess needs, implement suitable measures, document everything and refresh regularly.
  • Supervision: national authorities in each EU member state; in Germany, for example, the Federal Network Agency (Bundesnetzagentur).

Old and new wording compared

The Digital Omnibus on AI is Regulation (EU) 2026/1744. It was published in the Official Journal on 24 July 2026 and has applied since 27 July 2026. It replaced Article 4 in full.

Version until 26 July 2026Version since 27 July 2026
Obligationtake measures to ensure, to their best extent, a sufficient level of AI literacy of their stafftake measures to support the development of AI literacy of their staff
Goalsufficient level of literacysupporting development, explicitly no guaranteed level per person
Persons coveredstaff and other persons dealing with AI systems on their behalfunchanged
To be taken into accountprior knowledge, experience, education, context of use, persons affectedunchanged
New–the Commission and member states support companies, especially SMEs, for example with practical examples

In short: an obligation of result has become an obligation to take suitable measures. That takes the pressure off, because nobody has to prove any more that each individual has reached a specific level.

What has stayed the same

  • The obligation remains. Article 4 has applied since 2 February 2025 and has not been abolished.
  • The group affected is the same. It covers providers and deployers of AI systems. A deployer is any company that uses an AI system professionally. This also includes freelancers and service providers working with AI on your behalf.
  • The measures must be appropriate. Staff’s prior knowledge, the purpose of use and the people affected by the results still matter. Anyone using AI to select job applicants needs more than someone drafting social media posts.
  • There is supervision. Each member state has designated national authorities. In Germany, the Federal Network Agency has been responsible since the end of July 2026.

The AI Act does not provide for a specific fine for breaches of Article 4. The German Federal Network Agency points out, however, that a lack of AI literacy can be regarded as a breach of the duty of care, especially if it causes damage. This concerns, for example, liability towards customers.

Recommendations of a supervisory authority

The German Federal Network Agency describes three cornerstones on its AI literacy page. They are a useful guide in other EU countries too:

  1. Assess needs: who works with which AI systems, for what purpose, and what are the opportunities and risks?
  2. Design measures: clarify your own role, take prior knowledge and tasks into account, consider the context and risk of the system, define goals and responsibilities.
  3. Document, review, refresh: record measures, content, participants and period, review them regularly and sharpen them where necessary.

For content, it suggests a step-by-step structure: first, basics of data and AI with opportunities and risks; then your own role and the legal classification; finally, role-specific deep dives. The regulation does not prescribe a particular format. Self-study courses, workshops or multi-stage programmes are all possible, internal or external.

What “measures” mean in practice for SMEs

Most small and medium-sized businesses use AI for texts, research, summaries, images or analyses. For this everyday use, a lean solution that distinguishes by role is often enough:

RoleSuitable measureTypical content
Everyone who uses AIshort basic training, 2 to 4 hourshow language models work, recognising errors and invented facts, not entering confidential data, checking results
Teams with fixed AI workflows, e.g. marketing or salespractical workshop on your own toolswriting good instructions, checking sources, labelling AI content, copyright
Those responsible for an AI tool, e.g. IT or managementdeep dive on selection, settings and contractsdata protection and data processing agreements, access rights, internal AI policy
External parties using AI on your behalfreference to your rules, proof in the contract or by confirmationyour AI policy, handling of customer data

A short internal AI policy also helps: which tools are allowed, which data must not be entered and who the contact person is. How to introduce AI step by step is explained in our article Introducing AI in your company: the first step. What needs to be labelled in AI content for marketing is covered in our article on the labelling obligation under Art. 50 (in German).

Documentation in four building blocks

The German Federal Network Agency recommends documenting measures well so that you can show at any time that you comply with Article 4. A simple spreadsheet is enough:

  1. Overview: which AI tools are used, by whom and for what?
  2. Measures: which training sessions, workshops or materials were there, with what content?
  3. Participation: who took part and when? A list of participants or a short note is sufficient.
  4. Review: when will it be refreshed, for example once a year or when a new tool is introduced?

A certificate is not required for this. Certificates of attendance are still useful because they make documentation easier.

Free support

  • EU AI Office: webinars, frequently asked questions and practical examples on AI literacy.
  • European Digital Innovation Hubs: advice, testing and information on funding in all EU countries.
  • National authorities: in Germany, for example, the AI Service Desk of the Federal Network Agency with guidance, FAQs and a contact form.
  • Online courses: for example the free course Elements of AI.

Frequently asked questions

Does Article 4 apply if we only use ChatGPT?

Yes. Anyone who uses an AI system professionally is a deployer within the meaning of the AI Act. This also applies to general tools such as ChatGPT, Copilot or Gemini. Purely personal, non-professional use is excluded.

Do our employees need a certificate?

No. Neither a certificate nor a specific training is required. What matters is that you take suitable measures and record them in a traceable way.

Are there fines if we do nothing?

The AI Act does not provide for a specific fine for Article 4. However, a lack of AI literacy can be regarded as a breach of the duty of care, for example if an AI error causes damage. In addition, there are obligations under other laws, such as the GDPR.

Is a one-off training enough?

As a start, yes; in the long run, probably not. AI tools change quickly. The German Federal Network Agency recommends reviewing and refreshing measures regularly.

Conclusion

The new version of Article 4 makes the AI literacy obligation more manageable: measures instead of a guaranteed level, no certificate, free choice of format. Companies using AI should still act. Find out who works with what, train according to role, record it in a simple spreadsheet and refresh it regularly. This benefits not only compliance but above all the quality of the results.

Would you like to get your team up to speed in a practical way? Our workshop “Introducing AI in your company” covers basics, practice and rules in half a day, with a certificate of attendance for your documentation. Or arrange a free initial consultation.

Sources (as of 17 November 2026): Regulation (EU) 2026/1744 (Digital Omnibus on AI), EUR-Lex; Bundesnetzagentur, KI-Kompetenz (in German); Interface, AI literacy policy brief; Modulos, EU AI Act Article 4: AI Literacy Training by Role; ainora, EU AI Act Article 4. Note: this article provides a general overview and does not constitute legal advice.