AI in Hiring · 7 min read

EU AI Act Article 4: AI Literacy at Work and What Employers Must Do

What does EU AI Act Article 4 on AI literacy require of employers after the Digital Omnibus? Who it covers, a step plan per role and what to document.

Door Ingmar van Maurik · Founder & CEO, Making Moves


Short answer

Article 4 of the AI Act requires organisations that provide or use AI systems to take measures to support the development of AI literacy among their staff and others who work with AI on their behalf. Since the Digital Omnibus on AI, you do not have to guarantee any specific level. There is also no mandatory certificate or test result. What does count: that you can show you are doing something that fits how AI is used in your organisation.

In practice that means: knowing who uses which AI, deciding per role what is needed, training in a targeted way and documenting what you did.

*This article is general information, not legal advice.*

What does Article 4 say now?

Article 4 has applied since 2 February 2025. In July 2026 the Digital Omnibus on AI, Regulation (EU) 2026/1744, was published in the Official Journal of the EU. It entered into force in late July 2026 and replaced the text of Article 4.

Original text (from 2 February 2025)Current text (after the Digital Omnibus)

|---|---|---|

Core dutyTake measures to ensure, to their best extent, a sufficient level of AI literacyTake measures to support the development of AI literacy LevelA sufficient level as the aimExplicitly no duty to guarantee any specific level for any individual WhoStaff and others operating and using AI systems on the organisation's behalfStaff and others dealing with the operation and use of AI systems on the organisation's behalf

The text now also states that the European Commission and the Member States support AI literacy, especially for SMEs.

Commentators describe this as a shift from an obligation of result to an obligation of effort. It is a relaxation, not an abolition. The duty to take measures remains.

Who does it apply to?

Article 4 applies to providers (whoever develops an AI system or places it on the market) and deployers (organisations that use an AI system professionally under their own authority). For most employers, the second role is the relevant one.

  • The risk level does not matter. Article 4 also applies to low risk AI, such as a chatbot or a writing assistant.
  • Generic tools count. If employees use ChatGPT type tools, Copilot or AI features in your CRM or ATS for their work, your organisation uses AI.
  • People outside the payroll too. Think of temporary staff, freelancers or consultants who work with AI on your behalf.
  • No size threshold. Small companies are covered as well. What is reasonable does depend on how much and how you use AI.
  • What does AI literacy mean in practice?

    The AI Act describes AI literacy as the skills, knowledge and understanding needed to deploy AI systems in an informed way and to be aware of the opportunities, the risks and the possible harm. What that requires in concrete terms differs a lot per role.

    RoleTypical useWhat this person needs to be able to do

    |---|---|---|

    General userOccasionally has a text summarised or rewrittenKnow which data must not go into a tool, check output, recognise limitations Power userWorks with AI daily in their own processesGive good instructions, test results systematically, know when AI is not suitable BuilderCreates workflows, integrations or own applicationsTest, detect errors and bias, set up data and access securely Decision makerChooses tools, approves use, leads teamsWeigh opportunities and risks, set rules, know which legislation applies

    A recruiter who uses AI to write job ads needs something different from a developer who connects a screening tool to your ATS. One level for everyone rarely fits.

    Step plan for employers

    1. Map AI use

    Which AI tools are used, by whom and for what? Ask teams directly as well. A lot of use is not purchased centrally (shadow AI).

    2. Define roles and the level needed

    Group employees into the four roles above. For each role, record what someone needs to know and be able to do at a minimum. Take context into account: AI in hiring or customer contact calls for more care than AI for internal notes.

    3. Measure the current level per role

    A baseline shows where the real gaps are. You can use a short questionnaire, practical tasks or a structured scan. The Making Moves Team AI Skills Scan is one way to get a baseline per role. The scan is currently in a pilot phase. Read how to assess AI skills to see which formats work and which do not.

    4. Train in a targeted way

    Use the results to choose per role: a short basic briefing for general users, hands on training for power users, deeper training for builders and decision makers.

    5. Document your measures

    Record what you did and why. See below for the minimum to document.

    6. Repeat

    AI tools and how people use them change quickly. Plan a fixed review, for example once a year, plus an extra check when new tools or new roles appear.

    Why measure if no level is required?

    The law does not require a level. Measuring still helps, for three reasons.

  • Evidence of effort. A baseline and a follow up measurement show that you took measures and that they match practice.
  • Targeted training budget. You do not need to give everyone the same course. You invest where the gaps are.
  • Lower risk. If you know where knowledge is missing, you can prevent mistakes with confidential data, incorrect output or misuse earlier.
  • So measuring is a means, not a legal requirement. A score is not a certificate either.

    What should you document?

  • An overview of the AI systems in use and what they are used for
  • The division into roles and the intended level per role
  • The results of the baseline (group level is often enough)
  • Which training or briefing each group received, and when
  • Your internal guidelines for AI use: what is allowed and what is not
  • When you will review the approach again
  • Handle individual scores with care. The GDPR applies to them as well.

    Common mistakes

  • One generic e-learning for everyone. It is too light for builders and too abstract for people who use AI every day.
  • Treating it as a checkbox. A webinar attendance list says little about what people can do. An obligation of effort calls for measures that fit your use.
  • Ignoring shadow AI. If you only look at tools IT has purchased, you miss a large part of actual use.
  • Thinking only of IT. HR, sales, marketing and management use AI just as much.
  • Frequently asked questions

    Is AI literacy mandatory?

    Yes, in the sense that Article 4 of the AI Act requires organisations to take measures to support the development of AI literacy. Since the Digital Omnibus you do not have to guarantee a specific level for individual employees.

    Do employees need a certificate?

    No. The law does not require a certificate or a specific test result. A measurement or training can help you demonstrate your effort.

    Does it apply to small businesses too?

    Yes. There is no exemption based on size. What is reasonable does depend on how much and how you use AI. The European Commission and the Member States support AI literacy, especially for SMEs.

    Does employees' use of ChatGPT count?

    Yes, if employees use such tools for their work. Article 4 makes no distinction by risk category.

    How does this differ from the rules for high risk AI in hiring?

    That is a separate regime. AI systems that analyse or evaluate candidates in recruitment and selection are high risk under Annex III of the AI Act. After the Digital Omnibus, those obligations apply from 2 December 2027. In addition, the GDPR applies already, for example Article 22 on decisions based solely on automated processing.

    How often should you review AI literacy?

    The law sets no period. A yearly review plus an extra check when new tools or roles appear is a practical starting point.

    Summary

  • After the Digital Omnibus, Article 4 is an obligation of effort: you support the development of AI literacy without guaranteeing a level.
  • It applies to every organisation that uses AI professionally, including ChatGPT type tools and regardless of risk.
  • No certificate required, but measures that fit your use.
  • Work per role: a general user, power user, builder and decision maker each need something different.
  • Measuring, targeted training and documenting make your effort demonstrable.
  • High risk AI in hiring is a separate regime from 2 December 2027. The GDPR applies now.
  • Want to know where your teams stand? Explore the Making Moves AI Skills assessment and the Team AI Skills Scan or get in touch.


    Book intake call · View our AI Hiring System