EN
Deutsch
Français
Español
Italiano
ΕλληνικάKey takeaways
- No certificate is legally required under the EU AI Act; internal AI training records suffice.
- Good records capture who was trained, when, on what content, how it was role tailored, and refresher dates.
- National authorities begin supervising AI literacy duties on 2 August 2026, so get evidence ready now.
- For a small company, one spreadsheet plus verifiable certificates is proportionate documentation.
The EU AI Act never asks you for a certificate, but a supervisor can still ask you for proof. AI training records are the internal documentation showing that your staff received AI literacy training under Article 4 of Regulation (EU) 2024/1689, which has applied since 2 February 2025. The European Commission's AI literacy Q&A confirms that internal records suffice and no certificate is legally required. From 2 August 2026, national authorities will supervise exactly this duty, so the practical question is what those records should contain.
Do you need a certificate to prove AI training compliance?
No, you do not need a certificate to prove AI training compliance; the European Commission's AI literacy Q&A states plainly that internal training records are sufficient. The same guidance adds a point that matters just as much: handing staff the instructions for use of an AI tool is not enough on its own, and training is the expected practical measure. The bar is not a diploma. It is credible evidence that real training happened.
The 2026 Digital Omnibus, adopted by Parliament on 16 June 2026 and by Council on 29 June 2026, softened Article 4 into an obligation of effort: companies must take appropriate measures to support the development of AI literacy rather than guarantee a result. The duty itself remains in force and was not postponed. In an obligation of effort, documentation is the effort made visible. If you can show who was trained, when, and on what, you have done what the rule asks. For the legal background, see our guide to Article 4 and the AI literacy obligation.
What should AI training records contain?
A useful AI training record answers five questions: who was trained, when, on what content, how the training matched the person's role, and when it will be refreshed. Article 3(56) of the EU AI Act defines AI literacy as the skills, knowledge and understanding that allow people to make an informed deployment of AI systems and to gain awareness of the opportunities, risks and possible harm of AI, so your records should show that the content covered risks, not just tool tips.
- Who: name, job role and department, because the obligation is role and context based.
- When: the completion date, plus dates of any refreshers.
- What: the topics covered, including how AI works and its limits such as hallucinations, data protection and confidentiality.
- Role tailoring: why the depth of training fit the person's technical knowledge, experience and the context they use AI in.
- Refreshers: when the record will be reviewed, for example when new tools arrive or rules change.
This mirrors the compliance recipe the Commission and legal advisers converge on: map where AI is used, run role based training that covers risks and limits, tailor depth to role, and keep internal records of who was trained, when and on what. Your AI Act documentation does not need to be long. It needs to be answerable.
What does a simple AI training record template look like?
One table with seven fields covers everything an authority is likely to ask about. Keep it in a shared spreadsheet, one row per employee, and update it when people join, change roles or complete refreshers.
| Field | What to record | Example entry |
|---|---|---|
| Employee and role | Name, job role, department | Accounts payable clerk, finance team |
| Date trained | Completion date | 14 September 2026 |
| Content covered | Topics, with risks named | AI foundations, safe use at work, prompting, hallucination and confidentiality risks |
| Role tailoring | How depth matched the role | Finance specific module with invoice and data handling scenarios |
| Assessment | Quiz result or completion confirmation | Passed final quiz |
| Evidence | Where the proof lives | Certificate code checkable online |
| Next review | Refresher date or trigger | Review when a new AI tool is adopted |
The EU AI Office publishes a living repository of AI literacy practices, first released in February 2025, with real examples of how organizations implement Article 4. If you want reassurance that a lightweight table is normal, that repository is the reference point.
What will a national authority ask to see after 2 August 2026?
From 2 August 2026, national market surveillance authorities supervise the AI literacy rules, and the most natural opening request is simple: show us where your staff use AI and how you trained them. Expect questions in this order: which AI systems are in use, what measures you took to build sufficient AI literacy, how those measures were tailored to roles and context, and what records prove it. A completed version of the table above answers all four.
Two facts keep this in perspective. First, Article 4 breaches do not carry the headline AI Act fines; penalties are set by each Member State and must be proportionate, and as of mid 2026 no company had been fined under Article 4. Second, national implementation is uneven: per the tracker at artificialintelligenceact.eu, roughly nine Member States had designated authorities by mid 2026, about twelve had pending proposals, and about six had none. Note that the Digital Omnibus postponed obligations for high risk systems to 2027 and 2028 but did not postpone Article 4 or the supervision start. If you are still weighing whether the duty applies to you at all, read our article on whether AI training is mandatory in the EU.
How do verifiable certificates simplify AI literacy records?
A certificate with a verifiable code turns proof into a lookup: instead of assembling attendance sheets and slide decks, you point to a code that anyone, including an authority, can check online. Learnery is a browser based AI training platform that certifies employees in around eight hours for 39 euros per seat as a one time payment. Every completed seminar ends in exactly that kind of certificate, and the admin dashboard tracks completions and produces reports, so your AI literacy records largely maintain themselves.
The structure of the training also documents role tailoring for you. Each seminar has twelve steps: the first three cover AI foundations, safe use at work and everyday prompting, step four adapts to your industry, and step five to the exact subsector and role. Your record can state precisely which role specific content each person completed, in any of six languages. You can create an account and invite staff by email in minutes.
How much documentation is proportionate for a 20 person company?
For a 20 person company, a one page note on where AI is used plus the record table above is proportionate; the EU AI Act sets no fixed curriculum, hour count or exam for sufficient AI literacy. Measures only need to fit your staff's knowledge and the context AI is used in. Do not build a compliance binder nobody reads. Map your AI use in an afternoon, train people at the depth their role needs, and keep the spreadsheet current.
The cost side is equally contained. With Learnery, 20 seats qualify for the volume discount above 10 seats, so 39 euros becomes 31.20 euros per seat, or 624 euros total as a one time payment, and each person is trained and certified in around eight hours. You can weigh that against the time your team already loses to AI questions with the free ROI calculator, or talk to sales about a rollout. Whichever tools you choose, start the record now: the duty has applied since February 2025 and supervision starts on 2 August 2026.