No. The AI Omnibus entered into force on 27 July 2026 and dropped only the prescribed minimum level for AI literacy from article 4; the obligation itself remains, and the transparency obligations of article 50 became applicable on 2 August 2026, with a limited transition until 2 December 2026 for the Article 50(2) machine-readable marking obligation applicable to certain generative-AI systems placed on the market before 2 August 2026.
According to the European Commission on the page AI Omnibus enters into force, the Omnibus is part of the digital omnibus package proposed on 19 November 2025. In our assessment, the practical consequence for organisations working with AI is that the emphasis shifts from a possibly measurable competence threshold towards measures that you must be able to substantiate internally per context and risk, while you make AI interaction and certain AI content recognisable from the applicable date. This article uses only the official Commission sources; the distinction between the concepts is our editorial interpretation.
What exactly did the AI Omnibus change about article 4 on AI literacy?
Article 4 of the AI Act originally became applicable on 2 February 2025. On the page about AI talent, skills and literacy, the Commission explains that the Digital Omnibus, which entered into force in mid-July 2026, amended article 4. After that amendment, AI literacy remains an obligation for providers and deployers, but no specific or ‘sufficient’ level for individuals is prescribed any more.
The AI Office FAQ AI Literacy - Questions & Answers confirms that providers and deployers must take measures that take into account knowledge, experience, education, the context of use and the persons to whom systems are applied. That same FAQ explicitly states that the obligation does not require you to measure the knowledge level of every employee and that certificates are not needed.
Do the transparency obligations of article 50 still apply and from when?
Yes. In Quick Facts: Transparency rules for AI systems, the Commission describes that article 50 introduces transparency rules that apply from 2 August 2026. These concern obligations relating to:
- informing people when they interact with an AI system;
- marking AI-generated or manipulated content;
- deepfakes;
- certain AI-generated text on matters of public interest.
The same page mentions a transition period until December 2026 for certain markings of generative AI systems placed on the market before 2 August 2026. In our assessment, that transition means the conclusion ‘all transparency obligations are immediately fully enforceable for every existing system’ would be too broad; so check the market date per system.
Which dates and legal consequences must I keep apart?
The confusion around ‘rolled back’ arises because three moments and three legal consequences get mixed up. Keep them separate:
- 2 February 2025 — article 4 (AI literacy) became applicable.
- 27 July 2026 — the AI Omnibus entered into force and dropped the prescribed minimum level for AI literacy; the obligation itself remained in place.
- 2 August 2026 — the transparency obligations of article 50 became applicable, with a limited transition until 2 December 2026 for the Article 50(2) machine-readable marking obligation applicable to certain generative-AI systems placed on the market before 2 August 2026.
Our editorial reading: the Omnibus has not withdrawn the article 50 transparency obligations but has removed the prescribed AI-literacy outcome and left organisations to take measures tailored to the relevant knowledge, experience, training, use context and affected persons. More context on the broader set of obligations can be found in the topic hub on the EU AI Act and compliance.
How do I substantiate AI literacy measures without a prescribed minimum level?
The sources give the standard, but not the exact form of documentation. The AI Office FAQ does mention that internal records of training or other accompanying initiatives can be kept. On that basis, this is — as editorial advice, not as a source fact — a workable control point:
- Describe, per AI system, which functions and persons involved work with it and what risk that entails.
- Link concrete measures to that risk profile: instruction, guidance or training, tailored to knowledge and context of use.
- Record why those measures are appropriate, so that you can explain the choice without referring to an established minimum level.
- Keep internal records of training and guidance provided; a certificate per employee is, according to the Commission, not required.
For work with sensitive information, it helps to connect that substantiation with setting up demonstrable control over AI use by lawyers and with a verification layer for work with sensitive information.
What still applies to deployers of high-risk AI?
The AI Office FAQ makes clear that the relaxation of article 4 stands apart from other obligations. For deployers of high-risk AI, appropriate training for human oversight continues to apply separately. That requirement therefore does not disappear through the dropping of the minimum level for general AI literacy.
In our assessment, this means that organisations with high-risk applications must be able to demonstrate two tracks side by side: appropriate, context-bound AI literacy measures and specific preparation for human oversight per high-risk decision. How you record that oversight is set out in the piece on demonstrating and logging human oversight of high-risk decisions.
According to the page on AI talent and literacy, the Commission and member states must provide support and practical examples, and the AI Board must develop recommendations with common objectives. Until those exist, in our assessment the burden of substantiating appropriateness lies with the organisation itself.
Sources and references
Sources: The article relies on official European Commission pages about the AI Omnibus, the transparency rules of article 50 and AI literacy under article 4.