# EU AI Act timeline

Every date on which part of the EU AI Act starts to apply, as amended in July 2026, and the transitional rules for systems already on the market.

Source: https://sted.digital/docs/standards/eu-ai-act/timeline · Last reviewed 17 October 2026 · Strand Standards and regulations

EU AI Act Article 113 sets when each part applies, and Article 111 adds transitional rules, both as amended by Regulation (EU) 2026/1744. Prohibitions and AI literacy have applied since 2 February 2025, general-purpose model rules since 2 August 2025, transparency from 2 August 2026, and high-risk rules from 2 December 2027 (Annex III) or 2 August 2028 (Annex I).

## The requirement in plain English

The Act entered into force on 1 August 2024, 20 days after it was published in the Official Journal on 12 July 2024. Entering into force started the clock; it applied nothing by itself. EU AI Act Article 113 sets one general date, 2 August 2026, and then four exceptions, in the [consolidated text on EUR-Lex](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02024R1689-20260727):

- **Earlier, from 2 February 2025**: Chapters I and II, so the definitions, the AI literacy duty (EU AI Act Article 4) and the prohibited practices (EU AI Act Article 5).
- **Earlier, from 2 August 2025**: the rules on notified bodies, general-purpose AI models, governance and penalties, except the Commission's power to fine model providers.
- **Later, from 2 December 2027 and 2 August 2028**: the high-risk classification, requirements and obligations, for Annex III uses and for Annex I products respectively.
- **New, from 27 July 2026**: the Act's amendments to other EU product laws.

Regulation (EU) 2026/1744 also gave two new prohibitions their own date, 2 December 2026. Everything without its own date, including the transparency duties (EU AI Act Article 50) and the model-provider fines (EU AI Act Article 101), applies from 2 August 2026.

Article 111 handles what was already on the market. A high-risk system placed on the market before its high-risk date is covered only if its design changes significantly after that date (EU AI Act Article 111(2)), but a system intended for public authorities must comply by 2 August 2030. A general-purpose model placed on the market before 2 August 2025 must comply by 2 August 2027 (EU AI Act Article 111(3)). A generative system placed before 2 August 2026 has until 2 December 2026 to mark its output (EU AI Act Article 111(4)). Components of the large EU information systems in Annex X have until 31 December 2030 (EU AI Act Article 111(1)).

| Date | What starts to apply | Where it is set |
|---|---|---|
| 1 August 2024 | Entry into force; no duty applies yet | Article 113 |
| 2 February 2025 | Definitions, AI literacy, prohibited practices | Article 113, point (a) |
| 2 August 2025 | General-purpose model rules, notified bodies, governance, penalties | Article 113, point (b) |
| 2 February 2026 | Deadline for the Commission's high-risk classification guidelines (a draft followed on 19 May 2026) | Article 6(5) |
| 27 July 2026 | Regulation (EU) 2026/1744 in force; new AI literacy wording; amendments to other EU laws | Article 113, point (d) |
| 2 August 2026 | General date: transparency duties, fines on model providers, everything not dated elsewhere | Article 113 |
| 2 December 2026 | New prohibitions on non-consensual intimate imagery and child sexual abuse material; marking deadline for generative systems already on the market | Article 113, point (a); Article 111(4) |
| 2 August 2027 | Older general-purpose models comply; each Member State has a regulatory sandbox running | Article 111(3); Article 57(1) |
| 2 December 2027 | High-risk rules for Annex III uses | Article 113, point (c) |
| 2 August 2028 | High-risk rules for AI in Annex I products, such as medical devices | Article 113, point (c) |
| 2 August 2030 | High-risk systems intended for public authorities comply | Article 111(2) |
| 31 December 2030 | Components of Annex X large-scale IT systems comply | Article 111(1) |

This is guidance to help you understand the requirement, not legal advice.

## What good looks like

A 30-person recruitment software company keeps one register of its AI systems. Each row carries the system's role (provider for its screening tool, deployer for a bought-in writing assistant), its tier, the date each version was first placed on the EU market, and the dates from the table that apply to it. The compliance lead reviews the register every quarter and whenever the Commission publishes guidance.

Because the screening tool sorts job applicants, it is an Annex III use, so the company plans backwards from 2 December 2027: conformity work starts in 2026, a quality management system and technical documentation by mid-2027, and an internal rehearsal of the conformity assessment before the date.

The writing assistant needed AI literacy measures from 2 February 2025, and its marketing team's AI-written press releases need a label from 2 August 2026 unless an editor reviews them. Each duty has an owner and a dated record.

Release dates are evidence, not trivia. The company keeps release notes per version so it can show which transitional rule covers which system.

## Evidence an auditor expects

- **Placing-on-market dates per system and version**: a market surveillance authority checks these first when you rely on Article 111.
- **A dated classification for each system**: shows which tier, and so which date, you planned to.
- **Prohibited-use screening**: a record that each system was checked against Article 5 by 2 February 2025, and again for the December 2026 additions.
- **AI literacy measures since 2 February 2025**: who was trained or briefed, on what, and when.
- **Design change log for legacy high-risk systems**: shows whether a change was significant enough to bring an older system under the high-risk rules.
- **Marking plan for generative systems**: how output is marked, and that it was in place by 2 December 2026 for systems launched before 2 August 2026.

## Common mistakes

### Planning to the old dates

Slide decks from 2025 still show 2 August 2026 for high-risk systems and 2 August 2027 for products. The amended Article 113 moved both. Check every internal plan against the consolidated text dated 27 July 2026.

### Planning to a proposal

Until Regulation (EU) 2026/1744 was adopted in July 2026, the new dates existed only as a proposal and then a provisional agreement. Dates in a proposal are not law until the amending regulation is published. Cite the Official Journal, and treat later Commission proposals the same way.

### Reading "applies from" as "nothing to do before"

The high-risk date is when a system must already conform. A technical file, a risk process and a conformity assessment take months, so work starts well before.

### Assuming a legacy system is exempt for good

An older high-risk system escapes only until its design changes significantly. Define in your change process what "significant" means for each system and record the call each time.

### Training material with the old AI literacy wording

The duty was reworded on 27 July 2026. Courses written in 2025 describe the original text. Update them; the [AI literacy page](/docs/standards/eu-ai-act/ai-literacy) explains the difference.

## How Sted helps

Sted treats the EU AI Act as applying when at least one of your AI systems is in EU scope and carries obligations: it is prohibited, high-risk through the product route (it embeds a medical device under EU MDR classed above Class I) or through an Annex III use you recorded, or it is a generative system under the Article 50 transparency duties; a general-purpose model you trained and that is in EU scope counts only when it carries systemic risk; in-scope systems that are all minimal-risk give applies in part; and a stored high-risk assessment overrides the derived tier.
Which standards apply to you, and a device's class, are worked out by Sted's rules from your recorded answers; the AI model does not decide them.

Sted works out whether the Act applies to each recorded system. It does not decide which transitional rule covers a system already on the market; the provider does.

## Related guides and terms

- [EU AI Act](/docs/standards/eu-ai-act): the hub, with who the Act reaches and who enforces it.
- [Risk classification](/docs/standards/eu-ai-act/risk-classification): the tier that decides which date applies to a system.
- [Obligations for high-risk systems](/docs/standards/eu-ai-act/high-risk-obligations): what must be in place by 2 December 2027 or 2 August 2028.
- [Transparency obligations](/docs/standards/eu-ai-act/transparency-obligations): the duties that apply from 2 August 2026.
- [General-purpose AI models](/docs/standards/eu-ai-act/general-purpose-ai-models): the model rules that have applied since 2 August 2025.
- [ISO 42001](/docs/standards/iso-42001): the management system many organisations use to run the work to these dates.

## Frequently asked questions

Q: Was the EU AI Act delayed?
A: Partly. Regulation (EU) 2026/1744, in force on 27 July 2026, moved the high-risk rules from 2 August 2026 to 2 December 2027 for Annex III uses and from 2 August 2027 to 2 August 2028 for systems in Annex I products. The prohibitions, the general-purpose model rules and the Article 50 transparency duties kept their dates.

Q: Do the high-risk rules apply to a system already on the market?
A: Only if it goes through a significant change in its design after the high-risk date that applies to it (EU AI Act Article 111(2)). A system intended for use by public authorities must comply by 2 August 2030 whatever happens. The prohibitions apply to every system, old or new.

Q: Which date applies to a chatbot we launched in 2025?
A: The Article 50 duty to tell people they are talking to AI applies from 2 August 2026. If it also generates synthetic text, audio, images or video, the machine-readable marking duty in Article 50(2) applies to it from 2 December 2026, because it was on the market before 2 August 2026 (EU AI Act Article 111(4)).