Skip to content

EU AI Act Article 4 after the omnibus: what AI literacy requires now

By Sam Rivera, Founder, SentinelPanda · October 24, 2026 · 3 min read · AI Governance

The obligation got softer. The reasons to train the people who use AI did not.

What the original text required

As adopted in 2024, Article 4 required providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy among their staff and anyone else operating or using AI systems on their behalf. It asked them to take account of people's technical knowledge, experience, education and training, the context in which the systems are used, and the people the systems are used on. It applied from 2 February 2025, to every AI system regardless of risk class.

What the omnibus changed

The Digital Omnibus on AI, Regulation (EU) 2026/1744, was published in the Official Journal on 24 July 2026 and entered into force on 27 July. It rewrote Article 4 so that the duty is to take measures to support the development of AI literacy, and it states explicitly that this does not require guaranteeing any particular level for any individual. A new paragraph tasks the Commission and Member States with supporting providers and deployers, particularly SMEs, including by publishing practical examples.

Commentators disagree on how to characterise the change — an obligation of result becoming one of means, or an obligation that simply lost its benchmark. Either way, check the consolidated text on EUR-Lex before relying on any paraphrase, including this one.

What did not change

  • Who it binds: providers and deployers.
  • What it covers: all AI systems, not only high-risk ones.
  • Article 26(2): deployers of high-risk systems must assign human oversight to people with the necessary competence, training and authority — a literacy requirement in substance, untouched by the amendment.
  • ISO 42001 clause 7.2, if you are certified or pursuing certification, still requires you to determine and ensure competence.

Why it is still worth doing properly

The practical risks Article 4 was aimed at have not gone anywhere: staff pasting confidential data into public tools, automation bias in decisions that affect people, and unsanctioned AI use nobody has inventoried. Enterprise customers increasingly ask about AI training in vendor questionnaires. And the amended article still requires measures — so a supervisory authority can still ask what yours are.

A proportionate programme

  • Inventory who uses which AI tools, and for what.
  • Everyone: acceptable use, what data may not go into which tools, and how to recognise a confidently wrong answer.
  • People who build with AI: evaluation, prompt injection, data governance.
  • People who oversee AI-assisted decisions about others: system limitations, bias, and explicit authority to override.
  • Short modules, refreshed when the tools change, with completion and content versions recorded — that record is your evidence of "measures".
AI acceptable use policy Shadow AI usage EU AI Act timeline

Run your compliance program in one workspace.