AI Act training: How companies will fulfill the AI training obligation from 2025

Since February 2, 2025, companies in the EU have been obliged to train their employees in the use of AI systems. This obligation arises from Article 4 of the EU Regulation on Artificial Intelligence (AI Regulation), the so-called AI Act. The regulation affects almost every company that uses AI tools – from ChatGPT in marketing to automated applicant management systems in HR.

But what does this training obligation mean in concrete terms? Which companies are affected? And how can the requirement be implemented in a practical and legally compliant manner? This article provides you with a fact-based overview without scaremongering – with specific recommendations for HR managers and compliance officers.

What is the AI Act training obligation under Article 4?

The EU Regulation on Artificial Intelligence (Regulation (EU) 2024/1689) has been in force since August 2024. Article 4 of the regulation regulates the so-called AI competence obligation and obliges providers and operators of AI systems to ensure that “their personnel and other persons involved in the operation and use of AI systems on their behalf have a sufficient level of AI competence.”

What is AI competence?

According to Article 3 No. 56 of the AI Regulation, AI competence includes

• Skills and knowledge to use AI systems competently
• Understanding the opportunities and risks of AI
• Awareness of potential damage that can be caused by AI

The training should take into account the technical knowledge, experience, training and specific operational context of the employees.

Source: EU AI Act, Articles 4 and 3(56)

Who is affected by the AI training obligation?

The training obligation has been in force since February 2, 2025 and affects all companies that use or develop AI systems – regardless of size or industry.

Affected are:

Providers: Companies that develop AI systems and put them into operation or launch them on the market under their own name.

Operators: Organizations that use AI systems on their own responsibility without having developed them themselves.

Practical examples:

• A company that uses ChatGPT, Claude or Gemini for customer communication
• HR departments that use CV parsers or matching algorithms in recruiting
• Marketing teams that use AI-powered content tools
• Finance departments with automated risk analyses

Important: Even the use of common AI tools such as Microsoft Copilot or ChatGPT in an operational context is covered by the regulation.

Sources: IHK Munich on the AI Act, Haufe: Employer obligation AI competence

What are the consequences of non-compliance?

Clarity is important here: a breach of the training obligation under Article 4 is currently not directly punishable by fines. Article 99 of the AI Regulation, which regulates the sanctions, does not specify any direct penalties for non-compliance with the training obligation.

What threatens instead?

1. Liability risks: If damage is caused by incorrect AI use, this could be considered a breach of the employer’s general duty of care – especially if no appropriate training can be demonstrated.

2. Indirect sanctions: Non-cooperation with supervisory authorities or a lack of evidence could have indirect consequences (Art. 99 para. 5 AI Regulation).

3. Reputational damage: Companies that fail to meet their compliance obligations risk losing the trust of customers and business partners.

4. Competitive disadvantages: Particularly in public tenders, a lack of compliance can lead to exclusion.

Important classification

The training obligation is formulated as a best practice requirement (“to the best of our ability”). The EU Commission confirmed on February 20, 2025 that there is no certification requirement. However, it recommends documenting the time and participants of the training courses for liability cases.

Sources: Härting Rechtsanwälte: AI regulation employee training, IHK Munich: AI Act Webinar Info

How can companies fulfill the training obligation?

The AI Regulation does not specify any concrete training content or formats. Companies therefore have room for maneuver when it comes to implementation. A structured approach is recommended:

5-step approach to implementation

Step 1: Create an AI inventory Systematically record all AI systems used in the company:

• Which tools are used? (ChatGPT, Copilot, recruiting software, etc.)
• In which departments?
• For what purposes?
• At what risk level?

→ Download free EU AI Act checklist for recruiting

Step 2: Carry out a needs analysis Identify which employees need to be trained and what level of knowledge they have:

• Who works directly with AI systems?
• What prior knowledge do you have?
• What are the risks in the respective area of application?

Step 3: Develop a training concept Define suitable training formats depending on the target group:

• Basic training for all employees with AI contact
• Specialist training for intensive AI users (e.g. HR, marketing)
• In-depth training for high-risk areas

Step 4: Conduct & document training

• Choose a suitable training format (internal, external, online, face-to-face)
• Document participation, content and time
• Create optional certificates of attendance as proof

Step 5: Continuous updating AI develops quickly. Schedule regular refresher courses and adapt training content to new developments.

Source: IHK Cologne: AI competence obligation

What content should an AI Act training course cover?

Sound basic AI training should cover the following topics:

Module 1: AI basics

• What is artificial intelligence?
• How do common AI systems (LLMs, ML) work?
• Difference between different types of AI
• Typical use cases in the corporate context

Module 2: The EU AI Act

• Structure and objective of the regulation
• Risk classification (minimal, limited, high, unacceptable)
• Prohibited AI practices under Article 5
• Obligations for providers and operators
• High-risk AI systems (e.g. in recruiting)

Module 3: Legal framework

• Data protection (GDPR) in connection with AI
• Copyright and AI-generated content
• Transparency obligations
• Liability issues

Module 4: Risks and limits of AI

• Bias and discrimination
• Hallucinations and unreliability
• Data protection risks
• Ethical issues

Module 5: Practical application

• Safe use of AI tools
• Prompting basics
• Quality control of AI outputs
• Handling sensitive data

Source: GVW Rechtsanwälte: AI training contents

Training options at a glance

Companies have various options for fulfilling the training obligation:

Option 1: Internal training

Advantages: Individually customizable, cheaper for large teams

Disadvantages: High preparation effort, no external expertise, no standardized certificate

Option 2: IHK and chamber training courses

Advantages: Established providers, regional availability

Disadvantages: Often more expensive (400-800 EUR), deadline-dependent Examples: IHK Cologne, IHK Munich, IHK Rhine-Neckar

Option 3: University and research institutions

Advantages: High technical depth, scientifically sound

Disadvantages: Usually more expensive (500-2000 EUR), longer course duration Examples: Fraunhofer Institute, ZHAW

Option 4: Specialized training providers

Advantages: Flexibility, online formats, often with certificate

Disadvantages: Quality varies greatly, prices vary greatly (99-1950 EUR)

Option 5: Legally certified online training courses

Advantages: Scalable, flexible in terms of time, legally compliant, certificate as proof

Disadvantages: Less interaction than face-to-face formats

Example: German Institute for Artificial Intelligence (DIIFKI)

The DIIF-KI AI Act training as a certified option

The German Institute for Artificial Intelligence offers an online training course certified by Dr. Arno Malcher, a specialist lawyer for IT law, which is specifically tailored to the requirements of Article 4 of the AI Regulation.

Scope of services

4 modules, 35 lessons:

1. AI basics (clearly explained for non-technicians)
2. EU AI Act in detail (legal requirements)
3. Overview of AI tools (practical applications)
4. Rights & ethics (GDPR, copyright, ethical aspects)

Special features:

• Multimedia: read-aloud function, videos, interactive exercises
• Practice-oriented: Examples from everyday business life
• Flexible: self-determined learning pace
• Legally compliant: accredited by a specialist lawyer for IT law
• Certificate: As proof of competence after completion

Pricing

• Individual licenses / microenterprises: EUR 299 per person
• SMEs (20-50 licenses): 229 EUR per person
• Enterprise: On request
• 10% discount with the code KOOKU10 on all booked courses

Legal confirmation by Dr. Arno Malcher: “The AI basics course from the German Institute for Artificial Intelligence meets the compliance requirements of the AI Regulation with regard to teaching AI skills and sensitizes your employees to legal and commercial risks when using AI.”

→ Further information on the DIIF-KI training course

(With the code KOOKU10 you will receive 10% off all booked courses, should you choose this course).

Special feature: High-risk AI in recruiting

This is particularly relevant for HR departments: AI systems in recruiting are considered high-risk AI according to the AI Act (Annex III of the regulation). This concerns:

• CV parser with automated preselection
• Matching algorithms for candidates
• Video interview evaluation systems
• Automated scorings or rankings

Additional obligations for high-risk AI in recruiting

• Check the tool provider’s CE certificate of conformity
• Transparency obligation: Applicants must be informed about the use of AI
• Human supervision: Final decisions must be made by people
• Bias testing: Checking training data for discrimination
• Documentation: Store results for at least 5 years
• Incident management: report serious incidents within 15 days

→ To the download: Free 14-point checklist: EU AI Act in recruiting

Source: EU AI Act, Annex III; Kooku: AI Act in Recruiting

Practical checklist: How to implement the training obligation

✅ Phase 1: Analysis (week 1-2)

• [ ] Inventory all AI systems in use
• [ ] Identify affected employees
• [ ] Determine risk classes of the systems used
• [ ] Determine current state of knowledge

✅ Phase 2: Planning (week 2-3)

• [ ] Define training requirements per target group
• [ ] Select training format (internal/external, online/present)
• [ ] Determine budget and time frame
• [ ] Clarify responsibilities

✅ Phase 3: Implementation (week 4-8)

• [ ] Commission a training provider or prepare internal training
• [ ] Invite and schedule employees
• [ ] Conduct training courses
• [ ] Document participation

✅ Phase 4: Documentation (ongoing)

• [ ] Issue certificates of participation
• [ ] Archive training documents
• [ ] File documentation for proof of compliance
• [ ] Implement AI policy in the company

✅ Phase 5: Continuity (semi-annual/annual)

• [ ] Plan refresher training
• [ ] Train new employees
• [ ] Update training content
• [ ] Evaluate effectiveness

Support with implementation

Do you need support with the implementation of AI Act-compliant processes in your recruiting or HR department?

Kooku Recruiting offers specialized consulting:

• AI inventory for HR processes
• Risk assessment of recruiting tools
• Process optimization for AI Act compliance
• Training concepts for HR teams
• Data-driven recruiting strategies

→ Learn more about Kooku Consulting

Conclusion: Act pragmatically and with legal certainty

The AI Act training obligation has been a reality since February 2025 – but there’s no need to panic. The regulation is deliberately designed to be flexible and gives companies room for maneuver when implementing it.

Important key points:

✅ Clarify who is affected: Take an inventory of your AI tools and identify affected employees.

✅ Train appropriately: Choose training formats that suit your organization – internally or externally.

✅ Document: Keep verifiable records of training courses to minimize liability risks.

✅ Ensure continuity: Schedule regular updates.

✅ Legal certainty: In the case of high-risk AI (e.g. recruiting), training should be legally sound.

The training obligation is more than just a compliance requirement – it is an opportunity to make your employees fit for the future of AI and minimize risks at the same time.

Next steps

1. download free checklist → EU AI Act Checklist for Recruiting (14 points + RACI matrix)

2nd training course → DIIF-KI KI basic course (accredited by the specialist lawyer)

3. request advice → Kooku Consulting: AI Act compliance in the HR sector

About Kooku Recruiting

This article was created by Kooku Recruiting, your partner for data-driven, AI-supported and legally compliant recruiting. With over 10 years of experience and more than 200 clients, we support companies in optimizing their HR processes and making them future-proof.