Fehler 1
You are falling for providers who advertise with an alleged „training obligation according to Article 4 of the AI Regulation“.

That sounds dramatic.
However, it is often too broad.

Article 4 does not require a specific mandatory course.
And no certificate either.

The Federal Network Agency writes on this: „The AI Regulation does not specify a particular format for measures. The format can range from self-learning programmes to workshops and training sessions, right through to multi-level further education programmes – depending on the specific needs and the concrete context. This can take place both within and outside the organisation.“

Companies must ensure their employees have an adequate level of AI proficiency. Appropriate to the role. Appropriate to the application. Appropriate to the risk.

Not every organisation needs the same measures.
Not everyone needs the same depth of understanding.
And not every offer that advertises compliance makes sense.

How to expose weak providers:
You are selling fear instead of need.
You talk about „duty“, but not about your specific use cases.
And they offer standard foils for everyone, even though Art. 4 of the AI Regulation does not require a standardised approach.

Incidentally, the word ‚obligation‘ does not even appear in Art. 4 of the AI Regulation.

Error 2:
Booking training courses without first determining the actual demand.

Many companies train aimlessly.

This would first have to be clarified:

Which AI tools are used at all?
From whom?
For which tasks?
What are the risks?
And what specific knowledge is missing?

Only then does further training become a sensible investment.

Instead of general knowledge, investments should be made in individual topics: For example,
how you should specifically improve data quality or how you should document AI usage.

Error 3:
Spending too much money on basic knowledge.

No expensive specialised training is required for getting started. There are good ways to introduce employees to AI in a solid and sometimes free manner.

Two examples:

1. appliedai-institute[de] offers free introductory courses on the EU AI Act. They provide initial orientation. If you want more, you can find in-depth courses there for a fee, as well as from Ben Hansen and Aleksandr Tjulkanov.

2. ki-campus[org] - a free learning platform for AI, supported by the Stifterverband für die Deutsche Wissenschaft. Wide range, high quality, created by first-class partners.

In my opinion, a clever way is this:
Establish a broad and cost-efficient basis.
And only then make targeted investments where it really becomes company-specific.

My conclusion:

Not all AI training is useful.
Not every reference to Art. 4 of the AI Regulation is serious.
And not all basic knowledge has to be purchased at a high price.

The better way is usually simple:

Clarify initial needs.
Then qualify specifically.
And only spend money where standard knowledge is no longer sufficient.

 

 

Author: Achim Korten, April 2026