Since 2 February 2025, the EU AI Act has required companies that use AI to support the AI literacy of their staff. The rule is in Article 4. Many managing directors and HR leads are now looking for an AI training, a course or an "AI licence" for their team. The question behind it is almost always the same: what exactly do we have to do, and how do we prove it?
This article answers both. It explains what Article 4 requires, who it applies to, what a useful training covers and how to document it. At the end, I describe how I run this as a workshop programme directly in your company.
What Article 4 of the AI Act requires
Article 4 requires providers and deployers of AI systems to take measures that support the AI literacy of their staff. This also covers people who work with AI on their behalf, such as freelancers or service providers. The measures should fit the technical knowledge, experience and training of those people and the context in which the AI is used.
The Digital Omnibus, Regulation (EU) 2026/1744, rewrote Article 4 in summer 2026. Before, companies had to ensure "a sufficient level" of AI literacy. Now they must take measures that support the development of AI literacy. No specific level of literacy is required for any individual employee.
That sounds like relief, but it isn't. The obligation still exists. A company that does nothing cannot show that it met it. And a team that does not know the limits of AI makes expensive mistakes: confidential data in the wrong tool, unchecked answers in customer emails, or wrong figures in a quote.
Does the obligation apply to my company?
Very likely, yes. Article 4 applies regardless of risk class. A deployer is any company that uses an AI system under its own authority in a professional context. It is enough for employees to use ChatGPT, Microsoft Copilot, Claude, Gemini or AI features in existing software for their work.
Company size does not matter. A business with 20 employees is covered too. The scope of the measures can, however, follow actual usage. A trade business that uses AI for quote texts needs a different training from a software company with coding agents.
The AI Act does not require a certificate or an exam. There is no official "AI licence" either. Offers under that name are trainings from private providers. What matters is that the measure fits your company and that you can document it.
What a good AI literacy training covers
A training under Article 4 should do more than explain what AI is. It should cover the tools your team actually uses and the mistakes that typically happen with them. These topics belong in almost every programme:
| Topic | Content |
|---|---|
| Basics | How language models work, what they do well and where their limits are |
| Typical errors | Spotting hallucinations, outdated knowledge and convincingly worded false statements |
| Data protection | Which data may go into which tool, GDPR and data processing agreements |
| Security | Prompt injection, handling credentials, risks of agents with tool access |
| AI Act | Risk classes, prohibited practices and transparency obligations at a glance |
| Responsibility | Human review, approvals and internal rules for AI use |
| Practice | Exercises with your own tools and real tasks from daily work |
The practical part makes the difference. Employees learn most when they solve a real task from their work with AI and review the result together. Afterwards, they use the tools more safely and more often.
Training by role instead of one size for all
Not every department needs the same thing. A training works better when it weights the content by role:
| Role | Focus |
|---|---|
| Management | Obligations, liability, AI strategy and approval processes |
| Sales and marketing | Reviewing texts, protecting customer data, transparency towards customers |
| Administration, HR and accounting | Data protection, personal data, cross-checking results |
| Engineering and IT | Coding agents, security, access rights and verification |
A short shared part for everyone creates a common understanding. The deeper work happens in smaller groups. That keeps the time each person spends manageable.
Documenting the measure
The AI Act does not prescribe a specific format for evidence. You should still record what you did. If something goes wrong, for example a data protection incident or an enquiry from the supervisory authority, clean documentation is your proof. One entry per training can look like this:
Measure: AI literacy workshop under Art. 4 AI Act
Date: 12 March 2026, 09:00–13:00
Format: In-house, on site
Audience: Sales and customer service (14 people)
Tools: Microsoft Copilot, ChatGPT Team
Content: Basics, hallucinations, data protection,
internal AI policy, practical exercises
Delivered by: Riccardo Lamanna
Attachments: Attendance list, slides, AI policy v1.2
Next session: Refresher in 12 months or when a new tool is introduced
This includes the attendance list, the training materials and an internal AI policy. The policy defines which tools are allowed, which data may go into them and who reviews AI results before they are passed on. New employees get the training during onboarding.
Online course or in-house workshop?
There are many online courses on AI literacy. They are cheap and quick to roll out. For general background knowledge they can be enough. But they do not know your tools, your data or your processes. The content stays abstract, and many employees simply click through.
An in-house workshop starts where your team works. We use your approved tools, discuss real cases and answer questions on the spot. The result is not just proof of attendance, but a team that uses AI safely in daily work. For most mid-sized companies, that is the better choice.
How the programme works with me
I offer AI literacy training as a compact programme for companies. It runs in four steps:
- Initial call (30 minutes): We clarify which AI tools your team uses, which departments are involved and what you have already put in place.
- Tailoring: I adapt the content to your tools, roles and industry. If you do not yet have an AI policy, we draft one.
- Workshop: Half a day or a full day, on site or online, with a shared part and in-depth sessions by role.
- Documentation: You receive the materials, the attendance list and a record of the measure for your files.
At Junikom, I ran this workshop as part of a larger engagement. I describe what that looked like in the Junikom case study. The programme works just as well on its own, without any further project.
Frequently asked questions about AI literacy
When does the obligation apply? Article 4 has applied since 2 February 2025. The version revised by the Digital Omnibus has applied since late July 2026.
Do we need a certificate? No. The AI Act does not require a certificate. What matters is a suitable measure and traceable documentation.
How long does a training take? For most teams, half a day for the basics plus in-depth sessions for individual roles is enough. New tools or new employees are a reason for a refresher.
Do all employees have to take part? Everyone who uses or operates AI systems at work. That includes freelancers and external service providers who work with AI on your behalf.
Does the training replace other obligations? No. Data protection under the GDPR and other AI Act obligations, for example for high-risk systems, apply in addition. The training does help your team meet them day to day.
Book an AI literacy training for your team
If you want to build and document your team's AI literacy under Article 4 of the AI Act, the best next step is a direct conversation. In a free 30-minute initial call, we clarify which tools you use, who should take part and what the workshop looks like for you. Book an initial call now.