The EU AI Act in 2026: which dates have passed and which have moved
The AI Act applies in steps. Three of its dates have passed, the AI Omnibus moved the two high-risk dates, and the duty that reaches the most companies is already in force.
Which dates apply, from the Commission's own timeline
The European Commission records that the Act “entered into force on 1 August 2024”. Its obligations then start to apply in steps, and the Commission's timeline gives five dates:
| Date | What applies from that date | Source |
|---|---|---|
| 2 February 2025 | Prohibited AI practices and AI literacy obligations | European Commission |
| 2 August 2025 | Governance rules and obligations for GPAI models | European Commission |
| 2 August 2026 | The Act becomes applicable | European Commission |
| 2 December 2027 | Rules for high-risk use cases in certain sensitive areas (Annex III), extended to this date | European Commission |
| 2 August 2028 | Rules for high-risk AI systems embedded into regulated products (Annex I), after an extended transition period | European Commission |
The first three dates have passed. The last two are extensions: they come from the AI Omnibus, which entered into force on 27 July 2026.
The AI Omnibus also sets 2 December 2026. From that date two new prohibitions in Article 5 apply, and providers of generative systems placed on the market before 2 August 2026 must mark their output as artificially generated by then.
AI literacy: the obligation most companies already have
Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff ... This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.
Regulation (EU) 2024/1689, Article 4(1), as replaced by Regulation (EU) 2026/1744
A company that buys an assistant and gives it to its staff is a deployer, so this Article 4 duty has been its own since 2 February 2025. The measures depend on the context of use.
What to write down first: an inventory
Before any classification argument, there has to be an inventory. For every AI system in use, record the purpose, the provider, the model, the data it processes, who uses it, where the processing happens, the risk classification, the contractual terms and the security controls. The list usually runs longer than expected, because features inside existing software also count.
Private deployment does not create an exemption
The Regulation attaches to the AI system, its intended purpose and the role of the organisation. Running the model on your own hardware does not change whether the use case is prohibited, high-risk or neither.
What a private deployment does change is the evidence. Where the processing happens, who can reach the model and which version is running become facts the organisation can state about its own site rather than facts it has to obtain from a supplier.
This timeline is still moving, so a classification decision should be checked against the current text rather than against a summary, including this one.
With Bastion
What Bastion changes
Bastion is a private AI system delivered as one sealed box that runs inside your building. One monthly fee covers the hardware, the model, the hardened operating system and support, and nothing your team types leaves the building.
Bastion does not make an AI system compliant and does not claim to. It makes several of the inventory fields answerable without a vendor questionnaire: the model and its version, the processing location, the access path and the update history.
The Article 4 duty stays with the deployer. A defined internal system makes it easier to discharge, because the training can describe one tool with one set of rules.
Questions this article answers
- Does the EU AI Act apply in 2026?
- Yes: the European Commission records that the Act entered into force on 1 August 2024 and became applicable on 2 August 2026. Prohibited practices and AI literacy obligations applied earlier, from 2 February 2025, and the GPAI rules from 2 August 2025.
- When do the high-risk rules apply?
- The Commission's current timeline places the Annex III high-risk rules at 2 December 2027, and high-risk AI embedded in regulated products at 2 August 2028. Both were extended by the AI Omnibus, which entered into force on 27 July 2026.
- Does running AI on our own servers exempt us from the AI Act?
- No: the Regulation attaches to the AI system, its intended purpose and the role of the organisation, not to the location of the hardware. What a private deployment changes is the evidence: model version, processing location and access path become facts about your own site.
On the record
- 1
- 2
EUR-Lex
Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligenceread 2026-09-22
- 3
Read next
- High-risk AI: what the Act classifies, and what it does notThe classification rules in the Regulation's own words, the four conditions that can take an Annex III system back out of the category, the current dates, and why moving the server changes nothing.
- AI literacy is already an obligation, and it is not a training courseWhat the Article says, what the Regulation's own recital says it is for, why the duty is role-based rather than uniform, and a five-module programme a small company can run without a consultant.