EU AI Act training: what Article 4 asks of employers

Any company whose people use AI at work has to support their AI literacy. Here is what the duty says since the July 2026 amendment, the dates that matter, how to match training to roles, and the records that show you did it.

In short

  • Article 4 has applied since 2 February 2025. National authorities supervise it from 2 August 2026.
  • It applies to providers and deployers. A company whose staff use AI tools at work is a deployer.
  • The duty is to take measures that support AI literacy, fitted to each person's role and use of AI.
  • No certificate is required. An internal record of who was trained, on what and when is enough.

What Article 4 says

Article 4 of the AI Act (Regulation (EU) 2024/1689) is the AI literacy article. Since the amendment that came into force on 27 July 2026 (Regulation (EU) 2026/1744), it asks providers and deployers of AI systems to take measures that support the development of AI literacy among their staff and the other people who operate or use AI systems on their behalf.

Those measures should take into account each person's technical knowledge, experience, education and training, and the context the AI is used in, including the people it is used on. The amended text also says a company doesn't have to guarantee a particular level of literacy for any one person.

So the duty is about the programme. You decide what each group needs, you deliver it, and you keep the evidence that it happened.

Who it applies to

Providers build AI systems or place them on the market. Deployers use them under their own authority. A company that gives its staff an AI assistant, uses AI features in its HR or finance software, or lets teams use AI tools for customer work is a deployer.

The people in scope are your staff and anyone else who operates or uses AI systems on your behalf, such as contractors and agency workers. People who only meet AI as customers aren't covered by Article 4.

Key dates

DateWhat happens
2 Feb 2025Article 4 AI literacy and the bans on prohibited practices apply.
27 Jul 2026The amending regulation comes into force with the new Article 4 wording.
2 Aug 2026National market surveillance authorities supervise and enforce Article 4. Transparency duties under Article 50 apply.
2 Dec 2027Obligations for high-risk systems listed in Annex III apply, such as AI used in hiring.
2 Aug 2028Obligations for high-risk AI in products covered by Annex I apply.

Training that fits the role

The Commission expects training to fit the role, the risk and the context. One course for everyone rarely does, and handing staff the instructions for an AI tool is generally not enough on its own. A common pattern is a foundation for everyone, with more depth for the people who decide about AI, build it, buy it or run it.

GroupWhat they needCourse
All staff and contractorsRecognise AI, protect data, check outputs, say when AI was used, and report concernsAI literacy and responsible use, 40 min
Leaders and AI ownersAssign accountability, keep an inventory, classify risk, approve controls, review evidenceAI governance for leaders, 60 min
Product, delivery and procurementDefine the purpose, assess suppliers, test, document and manage changeBuilding, buying and deploying AI, 90 min
Engineering and operationsSecure the AI lifecycle, monitor it, and handle incidentsSecure AI lifecycle and incident response, 90 min
HR and recruitmentSpot banned and high-risk uses in hiring and work decisionsAI in employment and HR, 60 min
Teams on high-risk systemsClassification, provider and deployer duties, human oversight, impact assessmentsHigh-risk AI and impact assessments, 120 min

Write down why each group gets what it gets. A short role assessment, kept with the training records, explains your choices to anyone who asks later. See the courses.

Records to keep

The Commission's guidance says a certificate isn't required and an internal record of the training is enough. A record that answers an auditor holds, for each person:

  • who they are and their role;
  • the course they took and its version, so you can show it was current;
  • the date they finished, the score and the pass mark;
  • why that course fits their role.

Keep the content of each version too, and a log of who assigned what and when. When the law or your tools change, a dated update for everyone already trained shows the programme kept up. EU Training keeps all of this, and the client admin exports it in one file.

High-risk systems

Article 4 sits beside a separate duty for high-risk systems. A deployer of a high-risk system must give human oversight to people with the competence, training and authority it needs (Article 26(2)). If you use AI in hiring, credit decisions or another use listed in Annex III, the people overseeing it need training on that system and its risks, on top of AI literacy.

Questions

Is AI literacy training mandatory under the EU AI Act?

Article 4 asks every provider and deployer of AI systems to take measures that support the AI literacy of their staff and of others who use AI on their behalf. It has applied since 2 February 2025, and national authorities supervise it from 2 August 2026. A company whose people use AI tools at work is a deployer.

Do we need a certificate?

The law doesn't require one, and the Commission says an internal record of the training is enough. Every completion here comes with a certificate anyway, and the record behind it holds the course version, the score and the date.

Which of our people need training?

Anyone who uses or operates AI systems on your company's behalf, contractors included. The depth should match the role: AI literacy for everyone, more for people who lead, build, buy or run AI, and specialist courses for HR teams using AI in hiring and for teams on high-risk systems.

What happens when the law changes?

We publish a new version of the affected course. People trained before get the short yearly update in place of repeating the full course, and each record keeps the version the person passed.

What happens if we do nothing?

Penalties for Article 4 are set in each country's law, and authorities have supervised it since August 2026. The more common cost comes after an incident involving AI: without a record of training, it's hard to show the company took the measures the law asks for.

Is there separate training for high-risk AI systems?

Yes. Deployers of high-risk systems must also assign human oversight to people with the competence, training and authority it needs (Article 26(2)). That duty sits beside Article 4, and the high-risk course covers it.

Get the training and the records in place

We map your roles to the courses, bring your people in, and keep the record for every person. A 30-minute call shows you how.