If anyone in your company uses AI, the duty already applies.
Article 4 requires employers to ensure a sufficient level of AI literacy among the people who work with AI. It has applied since February 2025, and not only to technology companies.
We are not a law firm. We deliver the training, the internal policy and the documentation; a legal assessment of your specific systems belongs with a lawyer.
Three things you have to be able to show.
An inspection does not ask whether you hold a certificate. It asks what measures you took and whether you can evidence them.
The duty falls on providers and deployers alike — which includes a company that merely uses AI. It is enough that someone on the team opens ChatGPT, Copilot or Gemini for work.
If you have never told people what must not go into an AI tool, the duty is not met today — however well they happen to use it.
The regulation asks for no certificate. It asks for measures you can show.
The dates are widely misreported.
The Digital Omnibus moved part of the original schedule during 2026, and a great deal of the coverage still has not caught up.
This is where things stand today. If someone tells you high-risk systems apply from August 2026, they are reading the old schedule.
- 2 Feb 2025in forceAI literacy (Art. 4)
- 2 Aug 2026in forceTransparency (Art. 50)
- 2 Dec 2026upcomingMachine marking of older systems
- 2 Dec 2027postponedHigh-risk systems (Annex III)2 Aug 20262 Dec 2027
The Annex III date was moved by the Digital Omnibus. Most coverage online still quotes the original.
Breaching the transparency duties carries fines up to 7.5 million euro or 1 % of worldwide turnover. For high-risk systems the ceiling is 15 million or 3 %.
Training, a policy, and the paperwork behind both.
AI training for teams
Practical training on your own processes — what AI handles, where it lies, and what must never go into it.
An internal AI policy
Usage rules written for your company, not a downloaded template with someone else's name in the header.
Proof of training
Attendees, content and date — in a form you can hand to an inspector.
A follow-up for the advanced
Once the team has the basics, we build on it with tool-specific training and automation.
Common questions
Who in the company does it cover?
People who encounter AI as part of their work — which is not only IT. Typically marketing, having copy written and images generated; sales and support, where AI drafts replies to customers; HR, where it helps screen CVs; finance, where it pulls figures out of documents; and engineering, where an assistant writes code that ships.
Do we need an accredited certificate?
No. The regulation prescribes neither a format nor a certificate — it asks for a sufficient level of AI literacy and the ability to evidence the measures you took. Internal training, a written policy and an attendance record satisfy that. Anyone selling you a mandatory certificate is selling something the text does not contain.
We only use ChatGPT to write copy. Does this still apply?
Yes. The duty falls on the deployer — whoever uses the AI, not only whoever builds it. How much training is appropriate scales with risk, and writing marketing copy is not the same as screening CVs, but doing nothing is not an option in either case.
How often should we train?
The regulation sets no interval. In practice that means when new people join and whenever the tools change materially — with AI, roughly once a year. What matters more than frequency is being able to show when you last did it and what it covered.
Do we have to label AI-generated text and images?
Since August 2026, yes, where the output is published to the public and falls under Article 50 — chatbots, synthetic audio and video, and deepfakes above all. Systems placed on the market earlier have until December 2026 to add machine-readable marking.
Is it true that high-risk systems applied from August 2026?
Not any more. The Digital Omnibus moved the duties for stand-alone high-risk systems under Annex III to December 2027, and for AI embedded in regulated products to August 2028. A great deal of what is online still quotes the original date.
Tell us what your company does with AI.
Let us know which tools you use and in which teams. We reply within 24 hours with a proposed training scope and what your policy should cover.