The EU AI Act: timeline and preparation for industrial companies
This page is not legal advice
The dates and article numbers below come from the regulation in force; which class your system falls into is for your counsel to determine. Last checked: 27 October 2026.
Timeline
| Date | What starts applying |
|---|---|
| In force | Prohibited practices and general-purpose AI (GPAI) obligations |
| 2 August 2026 | Supervision and enforcement powers of the AI Office and national authorities; Article 50 transparency obligations |
| 2 December 2026 | End of the Article 50(2) transition for generative systems already on the market before August 2026 |
| 2 December 2027 | High-risk systems under Annex III (Article 6(2)) |
| 2 August 2028 | High-risk systems under Annex I (Article 6(1)) |
| 2 August 2030 | High-risk systems used by public authorities |
The high-risk dates were postponed by amending Regulation (EU) 2026/1744 (the Digital Omnibus, in force 27 July 2026). Sources still circulate with the old date — 2 August 2026 — so check which version of the text you are reading.
The distinction that matters for industry: Annex I or Annex III
AI that is a safety component of a product covered by product safety legislation — machinery, lifts, pressure equipment — falls under **Annex I**, and its date is 2 August 2028. Uses such as recruitment, credit scoring or critical infrastructure management fall under **Annex III**, dated 2 December 2027. For a manufacturer that distinction is eight months of preparation time.
What gets prepared technically
- Technical documentation and a data governance record: what the model was trained on, within which limits
- Logging: which input, at what time, produced which output
- Human oversight: where a decision can be stopped, and by whom
- Accuracy and robustness measurement: which test set, which threshold, which drift raises an alarm
- Transparency: does the user know they are talking to AI, is generated content marked
The part we do
This is not a service page. We build the technical counterpart of those five items: measurement and test sets are described under *Measuring AI Accuracy*, logging and monitoring under *Model Monitoring and LLMOps*, the data boundary and authorisation under *On-Premise and Private Cloud Setup*. Legal classification is not our work; building the evidence layer once the classification exists is.
Sources
- Regulation (EU) 2024/1689 — the AI Act
- Regulation (EU) 2026/1744 — the amending regulation that moved the high-risk dates (Digital Omnibus)
- European Commission, AI Act enforcement page
Last updated: