AI training In force since 2 Feb 2025 from CZK 12,000

AI literacy training under the AI Act

If anyone at your company uses AI at work, Article 4 of the EU AI Act requires you to take measures that support improving your people's AI literacy. The duty has applied since 2 February 2025 and an amendment in force since 27 July 2026 softened it considerably: you are now explicitly not required to guarantee a specific level of literacy for any individual. What remains is the duty to do something, and to be able to show it.

Who this applies to

The regulation speaks of providers and deployers of AI systems. A provider develops the system and puts it on the market, a deployer uses it in the course of its activity. Most companies fall into the second group, often without realising it.

Sales writes proposals and emails in ChatGPT or Copilot
Marketing generates copy and images, HR screens CVs
Finance runs invoices through a tool that has a model inside it
Support has a chatbot on the website, even one somebody else built

In all of these cases the company is a deployer and the duty applies. It does not matter that the tool was bought off the shelf, nor that only part of the team uses it.

What Article 4 says after the amendment

This is the text in force since 27 July 2026, when the amending regulation, the so called AI Omnibus, entered into force:

Providers and deployers of AI systems shall take measures to support the improvement of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee a specific level of AI literacy for individual persons.

Three things follow directly. The duty to act remains, what moved is the bar: instead of ensuring a sufficient level, you take measures that support improvement. The second sentence is new and carries the substance, a company need not guarantee a given level for any one person. And the duty does not stop at employees, it reaches contractors and suppliers who work with AI on the company's behalf.

The amendment also added two paragraphs: the Commission is to publish practical examples of compliance on a single information platform, and the Council is to adopt a recommendation setting common objectives. Once those land, what counts as enough will be clearer. The original text is in Regulation (EU) 2024/1689, the change came with Regulation (EU) 2026/1744.

What changed in July 2026

The amendment changes how the duty is worded, not whether it exists. The difference is worth knowing, especially when somebody sells you training by waving fines at you.

Before: take measures to ensure, to their best extent, a sufficient level of AI literacy.
Now: take measures to support the improvement of AI literacy, and explicitly no duty to guarantee a level for an individual.
New support: the Commission and Member States are to help, small and medium enterprises in particular, and practical examples are coming.
Unchanged: the duty still reaches anyone who provides AI systems or uses them in their activity, regardless of company size.

Dates that matter

1 August 2024. The AI Act entered into force, with its parts becoming applicable in stages.
2 February 2025. Article 4 started to apply, along with the ban on prohibited practices.
2 August 2025. Rules for general purpose models and the governance chapter kicked in.
24 and 27 July 2026. The amending regulation was published and entered into force. Article 4 got its new, milder wording.
2 August 2026. The next wave of applicability. Supervision is being built out behind the rules. Article 50 also becomes applicable on this date: anyone deploying an AI system meant to interact directly with people must design it so that a person knows they are talking to a machine. An ordinary website chatbot counts.
2 December 2027. Obligations for high-risk systems under Annex III begin to apply. The 2026 amendment pushed this date back by a year.
2 August 2028. Obligations for high-risk systems under Annex I, meaning AI built into products already covered by other EU legislation. Also pushed back by the amendment.

Anyone claiming that a missed training session carries a EUR 35 million fine is confusing Article 4 with the prohibited practices in Article 5. The penalty framework points elsewhere and Article 4 is supervised by national market surveillance authorities. Meeting this duty is not a big job, but you do need to be able to evidence it.

What the training should cover

The regulation prescribes no syllabus, and after the amendment it does not ask you to guarantee a level either. What follows from its purpose, and from Commission guidance, is that a person using AI should understand four things:

How the tool works and where it fails. That a model generates an answer rather than looking it up, and that it can invent a figure that sounds right.
How to check the output. What must always be verified, where spot checks are enough, and where you cannot proceed without a source.
What data must never go in. Personal data, trade secrets, anything under NDA. Plus a clear rule on which tools are approved.
When a human decides. Which steps may never pass without sign off, and who owns them.

How I run it

I am not a lawyer and I do not sell this as a legal service. I come at it from the other side: I build AI systems for companies, so I know where they break in practice, and that is exactly what people need to grasp. The training runs on your real tasks, not on generic examples.

Before the session we map which tools your people use and what for
Half a day to a full day, online or on site, groups of up to twelve
Hands on work on your own material, not slideware exercises
You get the outline, materials, a dated attendance record and a draft internal AI policy
A follow up call a few weeks later, because the real questions arrive during use

If you care more about getting value out of AI than about the duty, ordinary AI training and workshops is the page you want. The two can be combined into a single day.

What to keep on file

An inspection will not ask for a certificate, it will ask what you did. The sensible minimum:

The training outline and the date it took place
A list of the people trained and their roles
An internal AI policy, ideally signed or acknowledged
A list of approved tools and what may be put into them
A note on how new joiners are covered, so it is not a one off event

Parameters

half to full daydepending on team size and roles

from CZK 12,000group of up to 12

online or on siteCzechia, Slovakia, elsewhere in the EU by arrangement

FAQ

Since when do employers have to train staff on AI?+

Since 2 February 2025. Article 4 of the EU AI Act requires providers and deployers of AI systems to take measures supporting the improvement of AI literacy among staff and others working with AI on their behalf. The original wording spoke of ensuring a sufficient level; an amendment in force since 27 July 2026 softened it.

Does it apply to small companies too?+

Yes. Article 4 draws no line by company size or by the risk class of the system. What matters is whether someone at your company uses AI at work. A five person firm writing proposals in ChatGPT falls under the same duty as a corporation running its own model. The scope of what you do should match how you use AI, so a small firm does not need what a bank needs.

Do we need certificates?+

No. The regulation prescribes no certificate, exam or minimum duration, and since the amendment it explicitly does not require you to guarantee a level of knowledge for any individual. What you show is that you took measures. In practice that means the training outline, a dated attendance record and an internal AI policy.

What changed in summer 2026?+

Two things. Since 27 July 2026 the amending Regulation (EU) 2026/1744 gives Article 4 a milder wording: instead of ensuring a sufficient level it speaks of measures supporting improvement, and it explicitly drops any requirement to guarantee a level for an individual. And from 2 August 2026 the next wave of the AI Act applies, so supervision is being built out behind the rules.

Does this apply outside Czechia?+

Yes. It is a regulation, not a directive, so the same text applies in all EU member states without national transposition. What differs country by country is which authority supervises it. I run the training in English online for teams anywhere in the EU, and on site by arrangement.

What does AI literacy training cost?+

From CZK 12,000 for a half day workshop for a group of up to twelve, online or at your office. That covers an outline tailored to your field, materials, the attendance record and a draft internal AI policy. A firm price follows in writing after an intro call. Full pricing is on a separate page.

AI training & workshops, when it is about team output rather than the duty
AI automation, when AI should take over concrete work
Data, RAG & backend, when AI needs safe access to company documents
When an AI agent pays off, the payback maths without the marketing fog

Updated 4 August 2026 to reflect Regulation (EU) 2026/1744, which amends Article 4 with effect from 27 July 2026. This page sets out how I read the regulation as someone who builds AI systems. It is not legal advice. For contested cases, and for high risk systems in particular, ask a lawyer specialising in AI regulation.

Tell me how many people you are and which tools they use, and I will send you an outline and a price. If an hour and a written policy is all you need, I will say so.

Write to me, I reply within hours.