AI Act Article 4: what the Digital Omnibus changes for AI literacy

Since 2 February 2025, Article 4 of the AI Act (Regulation (EU) 2024/1689) has asked organisations that use AI to look after the "AI literacy" of their staff. On 24 July 2026 the Official Journal of the EU published Regulation (EU) 2026/1744, known as the "Digital Omnibus on AI". It entered into force on 27 July 2026 and rewrites Article 4.

Two contradictory messages have been circulating since: "the training obligation has gone" and "nothing has changed". Neither is accurate. Here is what the text says, what it means for a Belgian SME and what you can do without spending weeks on it.

What changed in Article 4

Before: ensure a sufficient level

In its 2024 version, Article 4 required providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy. That level had to reflect people's technical knowledge, experience and training, and the context in which the systems are used.

After: take measures to support

Regulation 2026/1744 replaces that wording. Providers and deployers must now take measures that support the development of AI literacy among their staff and the people who use AI systems on their behalf. The European Commission's FAQ, updated on 27 July 2026, states that the obligation does not require a specific level of AI literacy for any given person.

Legal commentators describe this as a move from an obligation of result to an obligation of effort. The text also provides that the Commission and Member States will support organisations, including by publishing practical examples, and that the European AI Board will adopt recommendations.

Who is covered

Article 4 applies to all providers and all deployers of AI systems, whatever their size and whatever the risk level of the system. A deployer is an organisation that uses an AI system under its own authority in a professional context.

In practice, an SME whose staff use ChatGPT, Microsoft Copilot, a meeting transcription tool or an AI feature built into its accounting software is a deployer. It is therefore covered, even if it develops no AI itself.

What stays in place

• The principle: take measures suited to the people and the context of use.
• Supervision: according to the Commission's FAQ, national market surveillance authorities, not the AI Office, supervise Article 4. Their supervision starts in August 2026.
• Penalties: Article 99 of the AI Act sets no specific fine for Article 4. Any penalties fall under national rules.
• High risk: if you deploy a high-risk AI system, the duty to assign human oversight to people with the necessary competence and training remains in full (Article 26). The Commission's FAQ says so explicitly.

The AI Act timeline after the Digital Omnibus

Regulation 2026/1744 also moved several deadlines. These are the dates that matter for a business, as published by the Commission and by the law firms Cuatrecasas, Abreu Advogados and Mayer Brown.

• 1 August 2024: AI Act enters into force
• 2 February 2025: Prohibited practices and Article 4 (AI literacy)
• 2 August 2025: Obligations for general-purpose AI (GPAI) models and governance
• 27 July 2026: Digital Omnibus (Regulation 2026/1744) enters into force
• 2 August 2026: General application, including the Article 50 transparency obligations
• 2 December 2026: Two new prohibitions (non-consensual sexual deepfakes, child sexual abuse material) and end of the transition period for marking generated content
• 2 December 2027: High-risk systems in Annex III (recruitment, education, credit, etc.), instead of 2 August 2026
• 2 August 2028: High-risk systems embedded in regulated products (Annex I)

The Annex III delay directly affects SMEs that use AI to screen job applications or assess staff. They get sixteen extra months, but the deadline is now fixed: it no longer depends on standards being published.

Five practical actions for an SME

According to the Commission's FAQ, no certificate is required and an internal record of training and initiatives is enough as evidence. The Commission suggests covering three topics: what AI is and how it works, the organisation's role (provider or deployer), and the risks of the systems in use and how to mitigate them.

The fifth action is to review everything once a year, or whenever a new tool arrives. AI features often appear through a routine update of existing software.

Why train anyway

The case is no longer only legal. According to the Acerta survey published on 17 May 2026 (2,000 workers and 600 employers), 21.4% of Belgian workers took AI training through their employer in the last twelve months, while 59.1% would like to. Among employers, 32% already offer AI training and 38.1% do not intend to.

The gap between what workers ask for and what employers offer is real. A short, documented training session meets Article 4. It also cuts the most common mistakes: confidential data pasted into a public tool, invented answers reused without checking, HR decisions taken without human review.