Article 4 of the EU AI Act applies to almost every organisation using AI and has applied since 2 February 2025. It’s also the obligation most companies have quietly ignored.
In 2026 the wording changed. The Digital Omnibus on AI, Regulation (EU) 2026/1744, which entered into force on 27 July 2026, softened Article 4. The duty is still there.
What Article 4 says now
Providers and deployers of AI systems must take measures to support the development of AI literacy of their staff and other people dealing with AI systems on their behalf. They must take into account those people’s technical knowledge, experience, education and training, and the context in which the AI is used. The amended text adds that this does not require them to guarantee any specific level of AI literacy for any individual.
The original text was stricter. It required organisations to ensure, to their best extent, a “sufficient level” of AI literacy among staff and others using AI on their behalf.
The Omnibus also gives the Commission and Member States a job: to support companies’ efforts, especially smaller ones. The Commission must also publish practical examples of compliance.
In plain English: if your people use AI at work, you must take real, reasonable steps to help them understand it. You don’t have to prove that every person reached a set standard.
Who it applies to
- Any organisation that uses AI professionally, not just companies building it. If your staff use an AI writing assistant, a chatbot or AI features in your software, you’re a deployer and Article 4 applies.
- Staff and “other persons”, which can include contractors, temps and service providers using AI on your behalf.
- Every risk level. Unlike most of the Act, this duty isn’t limited to high-risk AI.
What “AI literacy” means
The Act defines AI literacy as the skills, knowledge and understanding that allow people to make an informed deployment of AI systems and to be aware of the opportunities, risks and possible harms AI can cause.
There’s no exam and no mandatory certificate. What’s reasonable depends on the role:
- Everyone who uses AI tools should know what the tools can and can’t do, that outputs can be wrong, and what data must never be pasted in.
- People who rely on AI for decisions, such as HR, credit or customer service, need to understand the specific risks, such as bias and error, and when to override the output.
- People who build, buy or approve AI need a deeper understanding of risk, testing and the legal obligations.
A practical way to comply
You don’t need an expensive programme. A proportionate approach:
- Map who uses what. Your AI inventory tells you which teams use which tools.
- Set three levels of training: a short baseline for everyone, role-specific sessions for decision-makers, and deeper training for AI owners.
- Cover the essentials in the baseline: what generative AI is, why it makes things up, data you must never share (personal data, client secrets, credentials), your AI acceptable use policy, and how to report a problem.
- Make it part of onboarding, and refresh it when tools or rules change.
- Keep records: who was trained, on what and when. That’s your evidence if a regulator or customer asks.
Is there a fine for not doing it?
Article 4 doesn’t have its own fine in the Act. Penalties are set nationally, and enforcement depends on each member state. But the bigger risk is indirect: if something goes wrong with AI, for example a data leak or a discriminatory decision, a lack of training makes your position much worse with regulators, courts and customers.
The short version
If your staff use AI, train them in proportion to how they use it, and keep a record. A one-hour baseline session, role-specific guidance for high-stakes uses and a clear policy put you ahead of most organisations.
Official sources
- EU AI Act, Regulation (EU) 2024/1689: the official text on EUR-Lex
- Digital Omnibus on AI, Regulation (EU) 2026/1744: the July 2026 amendment that moved the high-risk dates
- AI Act Service Desk: the European Commission’s AI Act explorer, compliance checker and FAQs
More primary sources, with a short note on each, are on our Sources page.
Last reviewed 25 October 2026. Based on Articles 3(56) and 4 of Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744 (Digital Omnibus on AI). Updated 25 October 2026 to reflect the amended Article 4 wording. This article is general information, not legal advice.