TempusLex
← All posts
2026-04-29 · 2 min read

EU AI Act timeline: what applies when (2024–2027)

A milestone-by-milestone walkthrough of when each part of the AI Act becomes applicable, and what your team should do at each stage.

The EU AI Act entered into force on 1 August 2024, but its obligations land in waves. Here is the practical timeline most teams need to track.

February 2025 — Article 5 prohibitions

Six months after entry into force, the prohibitions in Article 5 became applicable. If your system involves any prohibited practice (subliminal manipulation, social scoring by public authorities, untargeted facial scraping, real-time remote biometric identification in public spaces by law enforcement, etc.), you must stop placing it on the market or putting it into service.

Action: run an Article 5 review now if you haven't already.

August 2025 — GPAI provider obligations

Twelve months after entry into force, the rules for General-Purpose AI providers became applicable. If you place a GPAI model on the EU market, you must:

  • maintain technical documentation of the model
  • provide downstream-provider information
  • implement a copyright policy for training data
  • publish a summary of training data

Models with systemic risk (≥10²⁵ FLOP training compute, or specifically designated) carry additional obligations including model evaluation and incident reporting.

August 2026 — High-risk obligations

Twenty-four months after entry into force, the bulk of obligations on high-risk AI systems become applicable. This is the longest single project on most teams' roadmaps:

  • risk management system (Article 9)
  • data governance (Article 10)
  • technical documentation per Annex IV (Article 11)
  • automatic logging (Article 12)
  • transparency toward deployers (Article 13)
  • human oversight (Article 14)
  • accuracy, robustness, cybersecurity (Article 15)
  • conformity assessment + CE marking (Article 43)
  • registration in the EU database (Article 49)

August 2027 — High-risk via product law

Thirty-six months after entry into force, Article 6(1) high-risk classifications kick in for AI that is a safety component of products covered by the harmonised legislation listed in Annex I. Examples: medical devices, machinery, toys.

What to do now

Most teams are best served by working backwards from August 2026. If you ship anything that might be high-risk, you need most of 2026 to get the documentation, conformity assessment, and CE marking in place. The further you are from the deadline, the cheaper the work; sliding into Q2 2026 with no Annex IV draft is a stress test you don't want.

*Documentation tool, not legal advice. Generated by AI Act Kit. Have qualified counsel review documents with legal effect.*

Procedural-deadline updates in your inbox

Reforms, landmark rulings, extraordinary stays. No spam, unsubscribe anytime.