Since February 2, 2025, a new reality has applied to companies using AI: Article 4 of the EU AI Act requires providers and deployers of AI systems to ensure a sufficient level of AI literacy among all employees working with these systems. The good news is that there is no rigid certification requirement, and since a legislative amendment in the summer of 2026, the regulation has become even more practical. However, the challenge remains: you need to be able to demonstrate that your teams know what they are doing.
This article explains what AI training for employees actually means, who is affected, and how you can implement the literacy requirement in practice—without unnecessary overhead.
What exactly does Article 4 of the EU AI Act require?
The legal basis is more precise than many assume, though it has recently changed. Originally, Article 4 required providers and deployers of AI systems to "ensure" a sufficient level of AI literacy—a relatively strict requirement. With the Digital Omnibus Regulation (EU) 2026/1744, which revised Article 4 on July 27, 2026, this was softened: providers and deployers are now required to take measures to support the development of their staff's AI literacy, rather than having to guarantee a specific level of competence for every single individual. The original obligation to achieve a specific result has thus become an obligation of effort. The legislature justifies this by stating that a strict guarantee requirement is not suitable for all companies and places an unnecessary burden on smaller businesses in particular.
What hasn't changed is that this still applies to everyone who develops, deploys, or operates AI systems on behalf of your organization—including external service providers and freelancers. The regulation still does not prescribe a fixed training format. There is no minimum number of hours, no mandatory curriculum, and no requirement for a specific certificate. What counts is proof that your measures are appropriate for the risk context and the role of the individual in question.
The literacy requirement applies regardless of the AI system's risk class. Whether you use ChatGPT for internal research or operate a high-risk AI system for recruitment, in both cases, you should ensure that users understand what they are doing. For high-risk systems, there is an additional, more specific requirement: Article 26 of the AI Act separately requires operators of such systems to ensure that their personnel are adequately trained and capable of exercising effective human oversight. This obligation exists in addition to the general literacy requirement under Article 4 and is much more concrete.
Important for urgency: Since August 2, 2026, the Federal Network Agency (Bundesnetzagentur) has been the central market surveillance authority for the AI Act in Germany and is now authorized to request documentation, audit systems, and demand corrective measures. What previously sounded rather abstract now has a concrete supervisory structure.
Looking for a complete overview of all compliance requirements and deadlines? Check out our guide on EU AI Act readiness to learn how to prepare your organization.
Who exactly does the AI training requirement apply to?
The training requirement targets three main groups. First, all companies that develop or provide AI systems themselves. This affects not only tech giants but also medium-sized companies that develop, for example, AI-supported production control or automated customer communication.
In addition, it applies to all organizations that use AI systems—regardless of industry or company size. If your marketing team uses an AI tool for content creation, if the HR department uses AI-supported application analysis, or if sales works with predictive analytics systems, you are affected. Finally, the requirement also includes authorities and public bodies that use AI systems—it is not limited to the private sector but applies wherever AI is used.
Within your organization, different roles are affected to varying degrees:
- Developers and technical teams: They need a deep understanding of AI architectures, model behavior, and potential biases
- Day-to-day users: A basic understanding of how the system works and its limitations is usually sufficient here
- Executives and decision-makers: They need to understand the strategic and legal implications
- External partners: Service providers and freelancers who operate AI systems for you are also subject to the training requirement
How do you fulfill the competency requirement in practice?
The legislator focuses on organizational responsibility rather than red tape. This gives you room to maneuver, but it also requires well-thought-out documentation. In concrete terms, you can combine several measures to meet the requirements.
Internal training and workshops are the classic approach. AI training for employees can take various forms: from half-day introductory workshops for new tools to multi-day intensive training sessions for development teams. It is important that the content is tailored to each specific role.
Online training and e-learning offer flexibility and scalability. There are now numerous providers of online AI training, both paid and free. Platforms like Coursera or edX offer specific AI courses that can be easily integrated into the workday. Whether you choose an AI training course with a certificate is up to you – what matters is documented participation and proof of competency.
In larger organizations, multiplier models work particularly well: You train internal AI experts who then pass on their knowledge to their respective departments. This saves costs and ensures practical knowledge transfer. Policies and guidelines are a useful supplement to formal training – develop internal documentation that describes typical use cases, best practices, and do's and don'ts for your specific AI systems. This makes the competency requirements tangible and provides your teams with practical guidance.
Documentation is always key. Keep track of who completed which training and when, what qualifications they hold, and how you ensure ongoing competency. A simple overview in your HR system is enough—the main thing is that you can prove you’ve met your duty of care if needed. Especially since the Federal Network Agency has been actively conducting audits since August 2026, this documentation is more valuable than ever.
How much does it cost to implement mandatory AI training?
Costs vary significantly depending on company size and the approach you choose. Comprehensive AI training programs from external providers typically cost between 500 and 2,000 euros per person, while more compact online formats can start as low as 80 to 300 euros per person. In-house workshops for entire teams usually range from 2,500 to 7,000 euros per day.
For smaller teams or those just getting started, there are also free AI training options for employees—many platforms offer basic courses at no charge. You’ll usually miss out on custom tailoring and certificates, but it’s a great way to gain initial experience. Alternatively, you can leverage internal resources: if you already have AI expertise in-house, you can develop your own training formats with a manageable time investment that fit your specific systems and processes perfectly.
In the long run, a hybrid approach is usually best: use external training for fundamentals and specialized topics, complemented by internal formats for system-specific details and regular updates.
What separates good AI training from bad?
Not every training program meets the requirements equally well. Look for the following criteria when selecting training providers or developing your own internal programs:
Practical application over theoretical overload: The training should address concrete use cases from your daily work. Abstract AI theory isn't very helpful if employees don't understand how to apply what they've learned.
Role-specific content: A developer needs different skills than someone who just uses AI tools. Good training programs differentiate based on the target audience.
Content relevance: AI is evolving rapidly. Training that only covers generic basics becomes outdated quickly. Make sure that current developments and new risks are also addressed.
Legal context: Solid AI training for companies also covers the legal framework – from data protection and the AI Act to industry-specific requirements.
Documentable learning outcomes: Whether it's a certificate of participation, a final test, or documented case studies, you need verifiable proof that the training was completed and knowledge was imparted.
"This isn't about dry legal compliance, but practical everyday value: Only teams that truly understand AI tools will use them productively while avoiding data leaks or costly missteps."
Conclusion: See the competence requirement as an opportunity
The AI training requirement under Article 4 of the EU AI Act might initially seem like just another compliance hurdle. In reality, it’s a sensible measure: companies that use AI without properly preparing their teams risk not only regulatory issues but also operational errors. And even though the summer 2026 reform softened the requirement from a strict obligation to a best-efforts basis, it doesn't change the fact that documented, well-thought-out training is in every company's best interest—especially now that the Federal Network Agency can actively conduct audits.
The best way to start: First, get an overview of which AI systems are being used in your organization and who is working with them. Use this to develop a tiered training concept that fits your resources. Start with the most critical systems and roles and expand the program step by step. Document your measures from the beginning—this saves effort later and creates legal certainty.
FAQ: Frequently asked questions about the AI training requirement
Who is required to undergo AI training?
The requirement applies to everyone who develops, operates, or uses AI systems on behalf of your organization. This includes your own employees as well as contracted service providers and freelancers, regardless of the AI system's risk class. Since the legislative change in July 2026, it is formally an obligation to provide support rather than a strict guarantee of outcome—but in practice, this changes little about the fact that you should document your measures.
How much does AI training cost?
Costs vary depending on the format and scope. Compact online courses start at around 80 to 300 euros per person, more comprehensive external programs usually range between 500 and 2,000 euros per person, and in-house workshops for entire teams typically cost between 2,500 and 7,000 euros per day. However, there are also free online courses for the basics. Internal training primarily requires an investment of time, but it can be tailored more precisely to your systems.
How do you train employees for AI?
It's best to combine different formats: external training for the basics, internal workshops for system-specific details, e-learning for flexibility, and train-the-trainer models for larger teams. It is important to document all measures and have a regular update concept in place.
What types of AI training are available?
The spectrum ranges from introductory courses on how AI works to specialized programs for developers, end users, or management. Formats include in-person training, online courses, certificate programs, and internal workshops. The key is to ensure the training matches the role and the context of use.
Do I need separate training for every AI tool?
No, but you should ensure that users understand the specific functions, limitations, and risks of every system being used. For similar tools, a basic training session plus a short introduction to the specific features is often enough. For high-risk or complex systems, dedicated training makes sense—this is also where the specific training requirements under Article 26 of the AI Act apply.
Posted by:
Jens is the co-founder of bakedwith, a boutique consultancy firm specialising in smart automation and AI. He helps medium-sized companies and corporations to optimise processes, reduce manual work and achieve growth through efficient workflows. Previously, he was Growth Lead at OMR and worked as a freelancer, helping numerous companies optimise their CRM and automation systems. Jens is passionate about combining growth and efficiency to enable teams to focus on what matters most.
.webp)
.webp)
%201.png)
