What changed in Article 4 of the EU AI Act in July 2026?
Regulation (EU) 2026/1744, the Digital Omnibus on AI, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It rewrote Article 4 so that providers and deployers must take measures to support the development of AI literacy among their staff and the people using AI on their behalf. The old wording, which asked for a sufficient level of AI literacy, is gone, and the article now states that it does not require providers or deployers to guarantee any specific level of AI literacy of any individual. The obligation itself is unchanged in scope and has applied since 2 February 2025.
Key facts at a glance
- Regulation: Regulation (EU) 2026/1744, amending Regulation (EU) 2024/1689 (the AI Act).
- Published: 24 July 2026. In force: 27 July 2026.
- Old duty: ensure, to their best extent, a sufficient level of AI literacy.
- New duty: take measures to support the development of AI literacy.
- Explicit clarification: no obligation to guarantee any specific level for any individual.
- Who is covered: every provider and deployer of AI systems, including their staff and other persons dealing with AI on their behalf.
- Enforcement: national market surveillance authorities, from 3 August 2026.
What did the old Article 4 wording ask for?
The original Article 4, in force from 2 February 2025, said providers and deployers "shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf". That phrasing caused two problems in practice. It read as an outcome you had to achieve for each person, and it invited a market of tests and certificates that promised to prove the outcome.
What does the new Article 4 wording ask for?
The rewritten article keeps the duty to take measures and drops the outcome language. You have to take measures that support AI literacy, calibrated to the systems in use, the people using them, and the context of use, and you do not have to guarantee that any named individual has reached any named standard. Three things follow from that.
- The evidence is a record of measures, not a set of scores. The European Commission's own Q&A on AI literacy says there is no need for a certificate, and that organisations can keep an internal record of trainings and other guiding initiatives.
- There is no obligation to measure or test employees. The same Q&A states that Article 4 does not entail an obligation to measure the knowledge of AI of employees. You can run a short check if it helps the teaching. Nothing requires it.
- Calibration is the judgement being asked of you. A programme that ignores which tools you use and who uses them has skipped the part the article actually describes.
What did the Digital Omnibus not change?
The Omnibus did not narrow who Article 4 applies to, did not remove the human-oversight training duty for deployers of high-risk systems, and did not defer the Article 50 transparency duties. Three points are worth stating plainly, because the word "relaxed" gets used loosely.
The scope did not narrow. If your staff use ChatGPT to draft emails, you are a deployer. The Commission's Q&A confirms that a company whose employees use ChatGPT for tasks like writing advertisement text or translating must comply with Article 4, and should inform staff about specific risks, hallucination being the example it gives. If your HR platform screens applicants with a ranking feature, you are a deployer of that feature whether or not you chose it. Contractors and agencies who deal with AI on your behalf are covered too.
The high-risk oversight duty was not touched, though its start date moved. Article 26 requires deployers of high-risk AI systems to assign human oversight to people with the necessary competence, training and authority. The Commission is explicit that relying on a system's instructions for use is not sufficient there, and its Omnibus note says that obligation remains in place. What the Omnibus did change is timing: the obligations for most high-risk systems listed in Annex III now apply from 2 December 2027, and for AI embedded in regulated products under Annex I from 2 August 2028.
Article 50 transparency duties were not deferred. Telling people they are talking to a chatbot, and labelling deepfakes, applied from 2 August 2026 as planned. Only providers of generative systems already on the market before that date got a grace period, to 2 December 2026, for machine-readable marking.
When did enforcement of Article 4 start?
The Commission's Q&A puts it in one sentence: Article 4 has applied since 2 February 2025, and the supervision and enforcement rules apply from 3 August 2026. In Ireland, the Regulation of Artificial Intelligence Act 2026 was signed into law on 21 July 2026. It sets up a distributed model in which existing sectoral regulators act as market surveillance authorities, coordinated by Oifig IS na hÉireann, the AI Office of Ireland, which was established with a first chief executive at the end of July 2026. Enforcement is proportionate and case by case, so the practical question a regulator asks is what your organisation did, and whether you can show it.
What should an Irish organisation do now?
If you have not yet taken a documented measure, the shortest defensible path is a short inventory of the AI in use, a session that covers the law and the specific risks of those tools for the people who use them, and a written record of both. That is the shape of our half-day workshop, and it follows the four steps the Commission's Q&A sets out for any Article 4 programme.
Sources
- European Commission, AI literacy: questions and answers. The certificate, record, testing, ChatGPT, enforcement date and four-step passages above are quoted from this page.
- Regulation (EU) 2026/1744 on EUR-Lex, the Digital Omnibus on AI, published 24 July 2026.
- Regulation (EU) 2024/1689, the AI Act, Article 4 and Article 26.
- Department of Enterprise, Tourism and Employment: AI Office of Ireland established, 30 July 2026.
This is general information about the EU AI Act, not legal advice. Enforcement is proportionate and case by case, and your sectoral regulator may issue guidance that applies to you.
