Yes, it is still mandatory. Since 2 February 2025, companies that use AI systems must take measures to support their staff’s AI literacy. The Digital Omnibus, in force since 27 July 2026, dropped the “sufficient level” wording but kept the duty. No course or certificate is required. Train by role, and keep a short record.
We are not lawyers. This is how we would set it up, and your counsel should check anything that turns on a legal reading.
Auf Deutsch: KI-Kompetenz nach Art. 4 KI-VO. Die Schulungspflicht gilt weiter, aber ohne festes Niveau und ohne Zertifikat.
Is AI literacy training mandatory under the EU AI Act?
The duty is mandatory. A formal training course is not. Article 4 applies to providers, who build AI systems, and to deployers, who use them at work. A company whose staff use ChatGPT or Copilot for summaries or spell checks is a deployer, as IHK Ostthüringen puts it. Company size and the risk class of the tool do not matter.
The Commission’s AI literacy Q&A is plain about the timeline. The duty has applied since 2 February 2025. National market surveillance authorities, not the AI Office, supervise and enforce it from 2 August 2026.
The same Q&A answers the most common case directly. A company whose employees use ChatGPT to write advertising text or translate it must comply. Staff “should be informed about the specific risks, for example hallucination.”
What did the Digital Omnibus change in Article 4?
The Digital Omnibus on AI rewrote Article 4. It was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. The Commission’s own pages say “mid-July”; we go by the Official Journal.
Before the Omnibus | Since 27 July 2026 | |
|---|---|---|
The duty | “take measures to ensure, to their best extent, a sufficient level of AI literacy” | “take measures to support the development of AI literacy” |
The level | A “sufficient level” | “does not require providers or deployers to guarantee any specific level of AI literacy of any individual” |
Who is covered | Staff and other persons using AI on your behalf | Unchanged |
Help from the EU | Not in the article | The Commission publishes practical examples; the AI Board adopts recommendations |
The old wording comes from the original AI Act. The new wording is on the AI Act Service Desk. The change is real but narrow. You no longer have to reach a level nobody defined. You still have to act, and you have to be able to show what you did.
Who needs AI literacy training?
Not every employee. Article 4 covers staff and “other persons dealing with the operation and use of AI systems on their behalf”. The Commission names a contractor, a service provider and a client as examples of those other persons. A freelancer who writes your newsletter with an AI tool would likely count.
IHK Ostthüringen draws the useful line. Occasional users of generative AI need a solid basic understanding. Heavy users, and people in sensitive areas such as HR or marketing, need more. People with a background in AI development are normally covered already, says the Commission. Even so, they should know your systems and the legal side.
For the other AI Act duties that apply now, see our overview for companies.
What should AI literacy training cover?
The Commission lists four steps a programme should consider at a minimum:
A general understanding of AI: what it is, how it works, which AI your company uses, and its opportunities and dangers.
Your company’s role: do you build AI systems, or use systems built by others?
The risk of the systems you use: what staff need to know, and which risks they must watch for.
Training built on that analysis, fitted to people’s existing knowledge and to the context of use.
Turn those steps into a small matrix. Three groups of people and four topics are enough for most companies of 20 to 500 people. The depth below is our suggestion, not a legal standard.
Topic | Everyone who uses AI | Heavy users and sensitive roles | People who set up or oversee AI |
|---|---|---|---|
What AI we use | The approved tools and what each is for | Same, plus the limits of each tool | Same, plus versions and settings |
Our role | We use AI built by others | Same | When a change could make us a provider |
Risks | Hallucination, personal data, trade secrets | Bias in HR or marketing output, copyright | Failure modes, logs, what to escalate |
What changes at work | Check output before it leaves you | A review step for every external text | Who approves, and how errors are reported |
Teach it on one real workflow, not on AI in general. A sales team learns more from checking ten AI-drafted offers than from a slide on neural networks. Walk the workflow in order: what data the tool may see, where it sits in the task, what it may do alone, and how mistakes get reported and fixed.
If you do not have written rules yet, start with a two-page AI policy template. The training then explains the policy.
What counts as proof of AI literacy?
You do not need a certificate. The Commission’s answer is short: “There is no need for a certificate. Organisations can keep an internal record of trainings and/or other guiding initiatives.” It also says no AI officer or governance board is required.
IHK Ostthüringen adds that the AI Act prescribes no set content, duration, exam or external certification. Briefings, guides and workshops can count as well as courses.
A record is still sensible. If something goes wrong, it is the easiest way to show what you did. The Bundesnetzagentur’s June 2025 guidance paper recommends documenting the type of measure, its scope and timing, and who took part. That paper predates the Omnibus and still uses the old “sufficient level” wording, so treat it as practical advice only.
One line per person is enough. Keep it in a spreadsheet next to your list of AI tools.
A caution on templates. The AI Office keeps a living repository of AI literacy practices from other organisations. The Service Desk says copying those practices “does not automatically grant presumption of compliance with Article 4”. Use them for ideas, then fit them to your own tools.
A two-week rollout for a 20 to 500 person company
This plan is our suggestion. Adjust it to your size and tools.
Days 1 and 2: list the AI in use. Ask each team lead which AI tools their people use, including free accounts. Mark the approved ones. This list is the start of your record.
Days 3 to 5: sort people into the three groups. Most will be in the first group. Name an owner for each tool.
Week 2: run short sessions per group. Use one real workflow per team. Hand out the policy and a one-page guide. Write the record lines the same day.
End of week 2: set the refresh. Put a date in the calendar, and a rule that a new tool or a big model change triggers a short update. Give people one place to report AI mistakes.
This fits as the first step of weeks 0 to 4 of an AI roadmap.
What about high-risk AI systems and GxP roles?
Here general awareness is not enough. Article 26(2) says deployers of high-risk AI systems “shall assign human oversight to natural persons who have the necessary competence, training and authority, as well as the necessary support.” The Commission adds that relying on the instructions for use is not sufficient.
For pharma and medtech sites, GMP rules on personnel already apply. In EMA’s Annex 22 workshop report, EU GMP Part I Chapter 2 on personnel qualification and training was cited as applicable to AI oversight. Training, qualification and knowledge testing were named as possible evidence for inspectors. If AI sits in a GMP workflow, the people who approve its output need role training, recorded in your quality system.
More on this in AI in GMP and on the people who approve AI output.
FAQ
Is AI literacy training mandatory under the EU AI Act?
The duty is mandatory for companies that build or use AI systems at work. It has applied since 2 February 2025. A formal course is not required: briefings, guides and workshops can also count, as long as they fit the roles and the tools.
Do we need an AI literacy certificate?
No. The European Commission says there is no need for a certificate, and that an internal record of trainings or other initiatives is enough. No AI officer is required either.
Who enforces Article 4, and what are the penalties?
National market surveillance authorities, from 2 August 2026. The AI Act sets no fine amount for Article 4 itself. Penalties come from national law, and the Commission says any sanction must be proportionate to the case.
Does Article 4 apply to Swiss companies?
Switzerland has no cross-sector AI law yet. The EU AI Act applies to Swiss companies that place AI systems on the EU market, or whose AI output is used in the EU, according to the Swiss KMU portal. The Commission says this reach also holds for Article 4. Ask your counsel how it applies to you.
How often should we refresh AI literacy training?
The AI Act sets no interval. A sensible rule is a yearly refresh, plus a short update whenever you add a tool or the model behind one changes. That rule is our suggestion.
Where to start
This week, list the AI tools your teams use and sort people into three groups. Next week, run one short session per group on a real workflow, and write one record line per person.
Training is easier when it is about one real workflow. Book a free workflow audit and we will pick the workflow your team should learn on.
Sources
AI Act Service Desk, Article 4: AI literacy, consolidated text as at 27 July 2026, accessed 5 November 2026.
European Commission, AI Literacy: Questions & Answers, accessed 5 November 2026.
Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal, 24 July 2026.
Regulation (EU) 2024/1689 (AI Act), Official Journal, 12 July 2024, original Article 4.
AI Act Service Desk, Resources: living repository of AI literacy practices, accessed 5 November 2026.
IHK Ostthüringen, KI-Kompetenz und Schulungspflichten, as of 5 August 2026.
Bundesnetzagentur, Hinweispapier: KI-Kompetenzen nach Artikel 4 KI-Verordnung, June 2025 (before the Omnibus).
AI Act Service Desk, Article 26: Obligations of deployers of high-risk AI systems, accessed 5 November 2026.
European Medicines Agency, Multistakeholder workshop on expert contributions to artificial intelligence guidance development (Annex 22), report EMA/156789/2026 on the workshop of 30 June 2026, page 7.
KMU-Portal (Swiss Confederation), Bis Anfang 2027 soll ein Vorentwurf zur KI-Regulierung ausgearbeitet werden, 23 September 2026.