Skip to content
Back to blog
Guides

AI literacy under Article 4: what education providers should do now

Article 4 of the EU AI Act, as amended in 2026: which AI literacy measures academies need for staff and lecturers, what to record and what companies ask for.

AI literacy under Article 4: what academies should do after the Digital Omnibus. Beside it, what Article 4 has required since 27 July 2026: measures that support AI literacy, no level guaranteed for individuals, no certificate required, records recommended.

The EU AI Act contains no training obligation in the literal sense. Article 4 requires measures, and since July 2026 that duty has been explicitly softened.

In many academies, AI is already part of everyday work. Administrators draft course descriptions with a chatbot, lecturers have practice questions suggested for their chapter, and participants learn with an AI tutor. In the German state of Baden-Württemberg, 73 percent of continuing education providers use AI tools, mainly for texts and teaching material (Weiterbildungsatlas BW 2025). That brings education providers under Article 4 of the EU AI Act. The European Commission explicitly names companies whose employees use ChatGPT to write advertising copy or translate texts (Commission Q&A).

Much of what comes up when you search for the "AI training obligation" dates from spring 2025. It describes a duty to "ensure" AI literacy. That version no longer applies. This article explains what Article 4 requires after the Digital Omnibus amendment, which measures make sense for staff and lecturers, what to document, and why the topic can also become a course offer to companies. It reflects the position in October 2026, uses Germany as its example for national enforcement and is not legal advice.

Key takeaways

  • Support is the standard: since 27 July 2026, providers and deployers of AI systems must support the development of AI literacy. Nobody has to guarantee a specific level.
  • Lecturers count too: the duty covers your own staff and other persons dealing with the operation and use of AI systems on the academy's behalf.
  • No certificate required: neither a certificate nor an AI officer is needed, and Germany's implementing act contains no fine for breaching Article 4.
  • Documentation still pays off: the German regulator recommends recording the type, scope and participants of your measures.
  • Companies are asking: in Germany, 66 percent of companies rate their employees' AI skills as low.

What Article 4 of the EU AI Act has required since July 2026

The regulation defines AI literacy precisely. It means "skills, knowledge and understanding" that allow providers, deployers and affected persons "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" (Article 3(56), Regulation (EU) 2024/1689).

In the original version, providers and deployers had to take measures "to ensure, to their best extent, a sufficient level of AI literacy" of their staff. The Digital Omnibus amending regulation (Regulation (EU) 2026/1744) replaced Article 4 in full. Paragraph 1 now 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 […]. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual."

Three things changed. Ensuring became supporting. A minimum level for individuals is explicitly ruled out. And the duty now has a counterpart: under the new paragraph 2, the Commission and the Member States support providers and deployers, in particular small and medium-sized enterprises, and the Commission publishes practical examples of how to comply. Under paragraph 3, the AI Board adopts recommendations with common objectives.

Recital 8 of the amending regulation gives the reason: stringent obligations to ensure a sufficient level of AI literacy would not suit all providers and deployers and weigh especially on smaller enterprises. Article 4 remains an obligation all the same. It has applied since 2 February 2025, and the new wording since the amending regulation entered into force on 27 July 2026.

Timeline of Article 4 of the EU AI Act: the duty applies from 2 February 2025, the softened wording enters into force on 27 July 2026, Germany's implementing act with the Federal Network Agency as supervisor on 29 July 2026, national authorities supervise from 2 August 2026, and high-risk obligations for education apply from 2 December 2027.

Who the duty covers in your academy

A deployer is anyone "using an AI system under its authority", except in a personal, non-professional activity (Article 3(4)). An academy that gives staff a chatbot or uses an AI tutor in its courses is a deployer. The risk level does not matter. Germany's Federal Network Agency (Bundesnetzagentur) stresses that Article 4 also applies to general-purpose chatbots, regardless of sector or organisation size (Bundesnetzagentur, in German).

Two groups are covered. First, your own staff, from management to course administration. Second, people who operate or use AI systems on the academy's behalf without being employees. The Commission gives contractors and service providers as examples, and so does the Bundesnetzagentur. For academies whose teaching relies heavily on freelance lecturers, it makes sense to include them as soon as they use AI as part of their teaching assignment.

Which measures make sense for staff and lecturers

The regulation prescribes no format. Self-study programmes, workshops, training sessions or multi-stage programmes all work, delivered in-house or externally (Bundesnetzagentur). The Commission notes, however, that in many cases simply relying on a system's instructions for use might be ineffective (Commission). The Bundesnetzagentur describes three building blocks for orientation.

1. Assess the need

Find out who in your academy uses which AI systems for what, and which risks and opportunities come with that. A short survey of staff and lecturers is enough to start. It often turns up tools that were never officially introduced.

2. Design measures by role

The Bundesnetzagentur suggests building up in stages: first a basic understanding of data and AI, then the organisation's own role and the risks of the systems it uses, and finally role-specific depth. For an academy, that can look like this:

RoleTypical AI useSuitable measure
Management and heads of programmesChoosing tools, planning coursesOverview of opportunities, risks and duties under the AI Act
Administration, sales, marketingTexts, emails, course descriptionsBasics, typical errors such as invented details, rules for personal data
LecturersPractice questions, case examples, feedbackBasics plus checking AI output against the curriculum and the current legal position
Lecturers who configure an AI tutorSetting rules, topics and answer behaviourIntroduction to the specific system, its limits and how to handle reported errors

A short usage policy on which tools may be used with which content is an obvious building block. It also helps protect your lecturers' materials.

3. Document, evaluate, refresh

AI systems and the way they are used change quickly. Check at regular intervals whether your measures reach their goal, and refresh them when a new tool arrives or tasks shift.

What to document and what is not required

The Commission is clear: there is no need for a certificate, and organisations can keep an internal record of trainings and other initiatives. Article 4 contains no obligation to measure employees' knowledge of AI (Commission). The Bundesnetzagentur lists explicitly what Article 4 does not require: guaranteed levels of AI literacy, formalised or standardised training, external certification and an AI officer.

Following the Bundesnetzagentur's guidance, you should still record:

  • The type of measure, such as a workshop, e-learning or an introduction to a system
  • Content and duration
  • Participants and date
  • The AI systems the measure relates to
  • The result of the evaluation and the planned refresher

A look at supervision shows why this is worth doing. Article 4 is enforced by national market surveillance authorities. In Germany, the Bundesnetzagentur has been the competent authority since 29 July 2026 (Section 2 KI-MIG, in German), and the fines provisions in Section 15 of that act do not mention Article 4. Other Member States set their own rules. According to the Commission, national authorities have supervised Article 4 since 2 August 2026 and could sanction infringements under national law, proportionately and case by case. That becomes more likely if an incident can be traced back to a lack of training. The Bundesnetzagentur adds that a lack of AI literacy can count as a breach of the duty of care, especially where damage occurs. Good records are therefore above all evidence for the day you need it.

If your academy uses an AI tutor

With an AI tutor, your academy deploys a system that participants use directly. The lecturers who set it up for their course therefore need more than the basics. They should know which sources the tutor answers from, where it can make mistakes and how to deal with a reported wrong answer. Our 30-minute test for AI tutors shows how to check this with one of your own chapters.

Ask your vendor which introduction it provides for lecturers, and add it to your record. acemate supports the start with onboarding workshops for lecturers, support and help pages.

The high-risk rules are a separate matter. Annex III point 3(b) covers AI that evaluates learning outcomes, including when those outcomes are used to steer the learning process. The obligations apply from 2 December 2027. According to the Commission's draft guidelines of May 2026 (paragraphs 222 to 224), this concerns systems whose evaluations feed into grades, qualifications or admissions. Tutoring systems with personalised feedback and adaptive practice that support ongoing learning are not to be classified as high-risk. That changes as soon as results feed into grades, certificates or admissions. The guidelines are still a draft.

Companies need AI literacy: what academies can offer

Article 4 applies to every company that uses AI, and that is now most of them. In Germany, 57 percent of companies with 20 or more employees use AI, up from 36 percent a year earlier. 70 percent train their employees in using AI, and 25 percent offer no training at all. At the same time, 66 percent of companies rate their employees' AI skills as low, and only 28 percent as high (Bitkom, September 2026, in German).

Bar chart from a Bitkom survey of 603 German companies with 20 or more employees, summer 2026: 11 percent train all employees in AI, 22 percent most of them, 37 percent selected employees and 25 percent none. 66 percent rate their employees' AI skills as low.

For academies that already serve companies, this is an obvious offer. What is needed are measures that fit the role and the context of use, as Article 4 asks: AI in accounting for the finance team, AI in customer service for sales, AI in HR for the people team. This is where an education provider with sector knowledge has the edge.

A credible offer has two features. Do not promise a legally required certification, because none exists. And expect free alternatives: the European Digital Innovation Hubs support small and medium-sized enterprises (Bundesnetzagentur), the EU-funded AI Skills Academy has been operating since 1 May 2026 (Commission), and the Commission's living repository shows more than 40 practice examples. A certificate of attendance from your academy can help your clients keep their own record.

You can build such a course on acemate from your own materials. The AI tutor answers from your course material and cites document and page, exercises with feedback consolidate what participants learn, and Analytics show you in aggregated form where the group needs support. Data is processed in the EU, and your content is not used to train AI models (AI control and data protection). Documenting the measure for your clients remains your task.

For your own academy, an hour is enough for a first step: survey the need, set one measure per role and start a record. Other AI questions for academies, from assessment to protecting your content, are covered in our article AI in continuing education. If you are considering an AI literacy course for your clients, the page for academies and professional education gives an overview.

Personal conversation

See what your courses are capable of

Personal learning support for participants and transparency for instructors, based on your own content and rules.

Arrange a no-obligation call