The text, before and after

Before, the original Article 4 of EU Regulation 2024/1689 required providers and deployers of AI systems to take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and of other people operating or using AI systems on their behalf, taking into account their technical knowledge, experience, education and training, the context of use and the people the systems are used on.

Now, under the text replaced by EU Regulation 2026/1744, providers and deployers take measures to support the development of AI literacy of the same people, with the same criteria. The new text adds that this obligation does not require them to ensure any specific level of AI literacy for any person. (Summary of the official text: see the sources at the end.)

What really changes

  1. From "ensure" to "support". The text no longer asks for a result, a sufficient level of skills, but for measures that support their development. In practice, the focus shifts from what people know to what the organization does to help them grow.
  2. No level to guarantee. The new text says so explicitly. The Regulation does not prescribe a number of hours, a minimum test or a certification.
  3. A role for the Commission and Member States. The new paragraph 2 asks the Commission and Member States to support providers and deployers, SMEs in particular; the Commission publishes practical examples of compliance on the single information platform provided for in Article 62. Paragraph 3 tasks the AI Board with adopting recommendations.

What doesn't change

  • The obligation remains. Anyone providing or using AI systems, schools and businesses included, must still take measures. It has not become optional.
  • The same proportionality criteria: technical knowledge, experience, education and training, context of use and the people the systems are used on.
  • Penalties look at the measures taken. Article 99(7)(g), not amended by the Omnibus, lists among the criteria for setting fines the operator's degree of responsibility, taking into account the technical and organisational measures it has implemented. Being able to show what you did remains the best protection.

The other dates the Omnibus moved

  • Annex III high-risk systems (including those used in education for admission, assessment of learning outcomes and monitoring of tests, and those used for recruitment): from August 2, 2026 to December 2, 2027.
  • Annex I high-risk systems (safety components of products already covered by other EU legislation): to August 2, 2028.
  • Transparency (Article 50): confirmed from August 2, 2026. Providers of generative AI systems placed on the market before that date have until December 2, 2026 to comply with the content-marking obligation.
  • New provisions on prohibited practices added to Article 5: applicable from December 2, 2026.

What to do now, in practice

  1. Update documents that quote the old text. AI plans, acceptable-use policies, training plans and internal notices often repeat "ensure a sufficient level": replace it with the text in force.
  2. Map who actually uses AI, with which tools and for what. It's the basis for proportionate measures.
  3. Scale measures to risk: basic training for people using generative tools to prepare materials, deeper training for those touching systems that affect assessments or decisions about people.
  4. Write clear usage rules: what can and cannot go into a chatbot, who checks outputs, how students, families or customers are informed.
  5. Keep track of what you do. The Regulation doesn't require a training register, but in an inspection what counts are the measures you can demonstrate.
The new Article 4 doesn't lower the bar: it moves it. What matters is no longer a level to guarantee, but the measures you took and can show.

Sources

Frequently asked questions

Did the Digital Omnibus abolish the AI literacy obligation?
No. Article 4 of the AI Act, as replaced by EU Regulation 2026/1744 from July 27, 2026, still requires providers and deployers of AI systems to take measures to support the development of AI literacy of their staff. The wording changes: they no longer have to ensure a sufficient level, and the text clarifies that no specific level must be ensured for any person.

When does the new Article 4 apply?
From July 27, 2026, the date of entry into force of EU Regulation 2026/1744 (Digital Omnibus), published in the Official Journal of the EU on July 24, 2026. The AI literacy obligation itself has applied since February 2, 2025.

Is a certificate or a minimum number of training hours required?
The Regulation prescribes neither a number of hours nor a certification. What matters is taking proportionate measures and being able to demonstrate them: certificates and registers are a useful way to document them, not a formal requirement.

When do the high-risk obligations apply to schools?
After the postponement decided by the Digital Omnibus, the obligations on Annex III high-risk systems, which include AI used for admission, assessment of learning outcomes and monitoring of tests, apply from December 2, 2027.