AI literacy under Article 4: training obligation

Since 2 February 2025, Art. 4 of the AI Regulation has required all companies using AI systems to take measures that support the AI competence of their employees. Since the Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026), it has been expressly stated that no certain level of competence of individual persons has to be guaranteed; the obligation to take measures remains. What is meant is not that everyone must be able to program, but that every person who works with AI knows its possibilities, limits and risks for their tasks. The regulation does not prescribe a specific certificate. Training adapted to the role and tool, clear rules and documentation are crucial.

As of October 2026. An overview of all obligations is provided in the article EU AI Act for SMEs.

What "AI competence" means according to the regulation

The Regulation describes AI literacy as the skills, knowledge and understanding necessary to use AI systems competently and to be aware of their opportunities and risks, including possible harms. Article 4 requires measures to support the development of this competence and specifies the benchmarks: technical knowledge, experience, education and training of persons, the context in which AI is used and the persons affected by it.

In practice, this means that an accountant who uses AI to read documents needs different knowledge than a sales employee who drafts offers or a manager who decides on the use of AI.

Who needs to be trained

Anyone who uses AI systems professionally, even occasionally and also free tools such as ChatGPT or translation services.
External persons who use AI on your behalf, such as freelancers or service providers. A contractual arrangement is often sufficient here.
Leaders who decide on the use and limits of AI.

If you don't work with AI at all, you don't need to be trained. In many companies, however, AI has long been used without the management knowing it; the article Shadow AI in the company describes how to find out.

What content a training course should cover

Building block	Contents	For whom
Basics	How language models work, why they make mistakes (hallucinations), what they can't do	all
Tools in the company	The approved tools, meaningful tasks, good inputs (prompts)	all users
Data protection and confidentiality	Which data does not belong in AI tools, difference between shared and private accounts	all
Audit and Responsibility	Check results, check sources, who is responsible	all
Legal context	Transparency obligations, high-risk areas, copyright	Executives, Responsible Persons
Subject-specific	Use in your own area, e.g. customer service, accounting, sales	per department

How to implement the training obligation

Inventory: Who uses which AI tools for what?
Set rules: A short AI guideline with allowed tools and taboo data is the basis of any training.
Plan training by role: A common basic part for everyone, building on this practical units per department with the real tools and tasks.
Document: Record participants, date, content and documents. This is your proof to supervisors, customers and in the event of liability.
Refresh: With new tools, new employees and at least once a year.

Does it have to be a certificate?

No. The AI Regulation does not require a specific format, exam or certificate. However, an off-the-shelf online course often only partially meets the requirement because it does not know your specific tools, data and processes. The most effective way is to provide practical training on your own use case, combined with written rules.

What happens if you don't train?

The AI Regulation does not provide for a separate fine for Art. 4. Nevertheless, a lack of training is not without consequences: If an employee makes a mistake with AI, such as providing incorrect information to a customer or a data protection violation, it is checked whether the company has fulfilled its obligations. Without verifiable training, you are then in a worse position. In addition, there is the practical benefit: trained teams use AI more safely and productively.

Funding for AI training

Counselling and accompanied implementation is eligible for funding: Through INQA coaching, the federal government covers 80% of the consulting costs of up to €1,200 per day for up to 12 days, i.e. up to €11,520 subsidy. In Mecklenburg-Western Pomerania and the other new federal states, BAFA consulting funding is also eligible at 80% until 31.12.2026 (50% in the old federal states). Funding is provided for consulting, not software. The funding check shows which program fits.

How INQA coaching works is described in the article INQA coaching process.

Frequently Asked Questions

Is AI training mandatory for employees?

Yes. Since 2 February 2025, companies using AI systems have been required to take measures to support the AI competence of their employees in accordance with Art. 4 of the AI Regulation. How this is done is not prescribed; training is the usual way.

Do employees need an AI certificate?

No. The regulation does not require a certificate or examination. It is important that the training fits the tasks and tools and is documented.

How often does AI training need to be repeated?

There is no fixed deadline. It makes sense to refresh when new tools are introduced, for new employees and at least once a year, because the tools change quickly.

Does the training obligation also apply if only ChatGPT is used?

Yes. The use of generally available tools such as ChatGPT for professional tasks is also the use of an AI system within the meaning of the regulation.

Is AI training funded?

Accompanied introduction and qualification can be funded at 80% via INQA coaching, with up to €11,520 subsidy. The funding check shows whether your project fits.

Next step

Our AI training courses are based on your tools and tasks and provide certificates of participation and documentation for proof at the same time. To coordinate, you can reach us via the contact page.

Source: Art. 4 AI Regulation (as amended by Regulation (EU) 2026/1744), Art. 99 AI Regulation.