All articlesEU AI Act Article 4: The AI Literacy Obligation That Affects Almost Every Company

EU AI Act Article 4: The AI Literacy Obligation That Affects Almost Every Company

Article 4 of the EU AI Act requires every company using AI to take measures that support the AI literacy of its staff. Since the Digital Omnibus (July 2026) it is clear: this is an obligation of effort – but still an obligation. Here is what it means in practice.

Christian Lechner

July 7, 2026·6 min

While half the economy prepares for the labeling obligation under Article 50 (from 2 August 2026), almost everyone overlooks an obligation that has been in force since 2 February 2025 – and that affects far more companies: the AI literacy obligation under Article 4 of the EU AI Act.

Update (October 2026): Regulation (EU) 2026/1744 (the "Digital Omnibus on AI", in force since 27 July 2026) rewrote Article 4. Instead of ensuring "a sufficient level of AI literacy", companies must now take measures to support the development of AI literacy. No one has to guarantee any specific level. This article has been updated to the new wording.

The bottom line up front: if your company uses AI – even just ChatGPT in marketing – you must take measures that support your staff's AI literacy. This is not a recommendation; it is applicable law – worded as an obligation of effort, not of result.

What Is Article 4 of the EU AI Act?

Article 4 requires providers and deployers of AI systems to take measures to support the development 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 of use.

Since 2026 it is stated explicitly: the obligation does not require guaranteeing any specific level of AI literacy of any individual. The Commission and the Member States are to support companies – SMEs in particular – including with practical examples.

In plain terms: you do not have to guarantee a literacy level for every person. But you do have to do something appropriate so the people involved understand what they are doing – the opportunities, risks, and limits of the AI they use.

What Changed – and What Did Not?

Until 26 July 2026	Since 27 July 2026
Obligation	ensure "a sufficient level of AI literacy"	take measures to support the development of AI literacy
Type of obligation	worded as a result	obligation of effort – no specific level to guarantee
Who is bound?	providers and deployers	unchanged: providers and deployers
Specific course or certificate required?	no	unchanged: no

Important: the new wording relaxes the result, not the whether. Doing nothing at all still does not satisfy Article 4.

Who Is Affected? Almost Everyone.

The key term is "deployer": any natural or legal person using an AI system in a professional context. You do not have to be an AI provider – simply using AI is enough.

Concretely affected, for example:

Marketing teams using ChatGPT, Jasper, or Copilot for content
Customer service with AI chatbots or AI-assisted ticketing
HR departments using AI to screen job applications
Engineering using GitHub Copilot or similar assistants
Design & e-commerce using Midjourney or DALL·E for images

In other words: there is hardly a modern company that is not affected.

Which Measures Fit?

The legislator deliberately prescribes no particular course and no certificate – what fits depends on role, prior knowledge, and context of use. A sensible measure should, however, cover at least:

How generative AI works – and where its limits are
Typical failure modes such as hallucinations and bias
Data protection and confidentiality when using AI tools
When human oversight and review are strictly required
Which labeling and transparency obligations (Art. 50) apply

Since When Does It Apply – and When Is It Enforced?

Date	Event
2 February 2025	Article 4 (AI literacy) in force
27 July 2026	Rewritten by Regulation (EU) 2026/1744: take measures instead of ensuring a level
2 August 2026	National supervisory authorities begin enforcement

So the obligation has applied continuously since February 2025 – what changed is its wording, not whether it applies.

Are There Fines?

Honestly: Article 4 has no dedicated fine tier in Article 99 – and the Omnibus did not change that. The obligation still has consequences: authorities can ask which measures you have taken, and AI literacy becomes relevant in:

Audits and regulatory reviews
Liability questions when an AI error causes harm
Due diligence by investors, customers, and partners

Precisely because no specific level is required, what matters is being able to show the measures you took – not because of a single penalty, but because that is exactly what gets asked in any scrutiny scenario.

What You Should Do Now

1. Take inventory Map where AI systems are used across your company – often in more places than you first think.

2. Identify roles Who works with which AI system? That determines which measure fits whom.

3. Train Give staff the fundamentals above – tailored to their role.

4. Document The most important step: keep verifiable records of the measures you took – who completed which content and when.

The Crux: Provability

Article 4 does not prescribe any particular documentation. But a measure you cannot evidence is hard to demonstrate when it counts. A documented record – who completed which content and when – is the simplest way to show you met your obligation. An email saying "please read up" is weak evidence.