The AI Act after the omnibus: timeline of obligations 2025–2030

Timeline

Date	What applies	Status
1 August 2024	Regulation (EU) 2024/1689 enters into force	applies
2 February 2025	Prohibited practices (Article 5), AI literacy (Article 4)	applies
27 July 2026	Regulation (EU) 2026/1744 (omnibus) enters into force	applies
2 August 2026	Transparency under Article 50 and most remaining provisions	applies
2 December 2026	New prohibitions; end of the transition period for machine-readable marking of outputs from systems placed on the market before 2 August 2026	coming
2 August 2027	National regulatory sandboxes	coming
2 December 2027	High-risk systems under Annex III	coming
2 August 2028	High-risk systems under Annex I (regulated products)	coming
2 August 2030	Adaptation of high-risk systems used by public authorities	coming

What the omnibus changed

Regulation (EU) 2026/1744 of 8 July 2026 (“Digital Omnibus on AI”) entered into force on 27 July 2026 — just before the original date of 2 August 2026, when the obligations for high-risk systems were due to start applying. The main changes:

• High risk postponed. Stand-alone high-risk systems (Annex III) from 2 December 2027, AI built into regulated products (Annex I) from 2 August 2028.
• AI literacy. The obligation changed from “ensuring a sufficient level of AI literacy” to an obligation to take measures to support it, such as training or information materials.
• New prohibitions. AI systems intended to create non-consensual intimate content and child sexual abuse material — from 2 December 2026.
• Narrower definition of a safety component. Systems used only for assistance, convenience, performance optimisation, process automation or quality control do not count as high risk if their failure does not endanger health or safety.
• Relief. Measures for SMEs (simplified technical documentation, priority access to sandboxes, a cap on fines) were extended to so-called small mid-caps as well.

What this means for an ordinary business

Most businesses in Slovakia do not produce AI but use it — a chatbot on the website, generating text or images, assistants in office software. For them, three things matter today:

1. Do not use prohibited practices (e.g. manipulative techniques or emotion recognition in the workplace, with exceptions).
2. Transparency — if a customer is talking to a chatbot, they must know it; deepfake content must be labelled.
3. AI literacy — people who work with AI should know what the tool does and where its limits are.

If you use AI in recruiting, assessing or dismissing employees, this is usually a high-risk area under Annex III. The obligations for deployers (e.g. human oversight, informing employees) arrive on 2 December 2027 — but preparation takes longer than it looks.

What is not yet certain

The separate Digital Omnibus package, which would change the GDPR, the cookie rules and the Data Act, has not been adopted. Do not confuse it with the AI omnibus, which has been adopted.