REGULATION (EU) 2024/1689 · OJEU L · 12 JULY 2024Official European Union source

Application schedule
of the European AI Regulation.

Regulation (EU) 2024/1689 comes into application through successive tiers between August 2024 and August 2028. Five milestones structure the compliance path: prohibited practices, general-purpose AI models, transparency, high-risk systems and regulated products.

Adoption13 Jun 2024

Entry into force1 Aug 2024

Full application2 Aug 2028

Reference180 articles · 13 annexes

Status as of August 10, 2026Phase 3 active

Obligations in force3 / 5 tiers applied

Sanctions activatedArt. 99 — €35M or 7% turnover

Next tierJ−478

Overview

Six application milestones, from August 2024 to August 2028.

Article 113 of the Regulation provides for progressive application. The omnibus proposal (under review) adds a staggered schedule for the high-risk deadline. Each tier triggers a new perimeter of binding obligations, backed by the sanctions provided for in Article 99.

PHASE 1Framing

PHASE 2Prohibitions

PHASE 3GPAI & governance

PHASE 4 · in progressTransparency

PHASE 5High risk

AUG 10, 2026

07/31/2024

Entry into force

Since 24 months

02/01/2025

Prohibited practices

Since 18 months

08/01/2025

GPAI & governance

Since 12 months

08/01/2026

Art. 50 transparency

Since 8 d

12/01/2027

Annex III high risk

Next · D−478

08/01/2028

Regulated products

4 oblig.

ApplicableNext deadlineUpcomingOmnibus schedule dates (March 2026)

Omnibus proposal — splitting the high-risk deadlineIN PROGRESS

Article 50 transparency obligations have applied since 2 August 2026. The calendar then shows 2 December 2027 for Annex III high-risk systems and 2 August 2028 for Annex I regulated products.

Track the file

Deadline details

Article by article, obligation by obligation.

31July2024Applicable for 24 months

ART. 113OJEU L · 12.07.2024

Entry into force of the Regulation

Twentieth day following publication in the Official Journal of the European Union. The Regulation becomes legally enforceable in all 27 Member States; the countdown to the next deadlines begins.

What changes that day

Art. 113Publication in the Official Journal, final enforceable textAll

Art. 64–70Creation of the European AI OfficeEU

Art. 56Launch of the codes of practice processGPAI

01February2025Applicable for 18 months

ART. 4–5CHAPTERS I & II

Prohibited practices and AI literacy

First operational tier. AI systems presenting an unacceptable risk are prohibited across the European territory. Any organisation using or developing AI systems must ensure sufficient literacy among its concerned personnel.

Applicable obligations

Art. 5(1)(a)Prohibition of subliminal or manipulative techniquesAll

Art. 5(1)(b)Prohibition of exploitation of vulnerabilitiesAll

Art. 5(1)(c)Prohibition of social scoringAll

Art. 5(1)(d)Restriction on real-time remote biometric identificationAll

Art. 5(1)(g)Prohibition of biometric categorisation based on sensitive dataAll

Art. 4AI Literacy obligation for all concerned personnelProviders

01August2025Applicable for 12 months

CHAPTER VART. 51–55, 70, 99

GPAI models and European governance

Providers of general-purpose AI models (GPAI) must technically document their models, comply with copyright law, and publish a summary of training data. National competent authorities are designated; the sanctions regime becomes fully operational.

Applicable obligations

Art. 53(1)(a)Technical documentation of GPAI models (Annex XI)GPAI

Art. 53(1)(c)Policy to comply with Union copyright lawGPAI

Art. 53(1)(d)Publication of a sufficiently detailed summary of training dataGPAI

Art. 55Enhanced obligations for systemic-risk modelsGPAI+

Art. 70Designation of national competent authoritiesMember states

Art. 99Applicable sanctions regime (€35M or 7% turnover)All

01August2026Applicable for 8 days

ART. 50CHAPTER IV

Transparency obligations (Article 50)

Applicable since 2 August 2026, Article 50 requires transparency for AI systems interacting with natural persons: disclosure, marking synthetic content and labelling deepfakes.

Applicable obligations, Transparency

Art. 50.1Information of persons interacting with an AI systemTransp.

Art. 50.2Marking of synthetic content (watermarking)Transp.

Art. 50.3Information during emotion recognition or biometric categorisationDeployers

Art. 50.4Labelling of publicly disseminated deepfakesDeployers

01December2027Next deadline · D−478

ANNEX IIIART. 6, 8–27CHAPTER III

High-risk systems (Annex III)

The most structuring deadline of the Regulation. All requirements applicable to AI systems classified as high risk come into application: risk management system, quality data, complete technical documentation, human oversight, robustness, EU declaration of conformity.

Applicable obligations, High risk

Art. 9Risk management system across the entire lifecycleProviders

Art. 10Data governance and representative training setsProviders

Art. 11Technical documentation compliant with Annex IV (14 sections)Providers

Art. 13Transparency and instructions for use to deployersProviders

Art. 14Effective human oversight throughout the lifecycleProviders

Art. 15Accuracy, robustness and cybersecurityProviders

Art. 17Documented quality management systemProviders

Art. 26Deployer obligations (usage log, oversight)Deployers

Art. 27Fundamental rights impact assessment (FRIA)Public dep.

Art. 47–48EU declaration of conformity and CE markingProviders

01August2028Upcoming · in 24 months

ANNEX IART. 6(1)

Systems embedded in regulated products

Last tier. High-risk obligations extend to AI systems that constitute a safety component of a product covered by Union harmonisation legislation: machinery, toys, medical devices, radio equipment, lifts, marine equipment.

Applicable obligations

Art. 6(1)Application of high-risk obligations to Annex I systemsProviders

Annex IIntegration into sector-specific conformity assessment proceduresProviders

Art. 11Technical documentation integrated into product documentationProviders

Art. 48CE marking covering both the product and the AI componentProviders

Official sources

Reference texts and implementing documents.

All dates and obligations cited are drawn from official texts published by European and French institutions. Direct links to primary sources.

EUR-Lex · European Union

Regulation (EU) 2024/1689, consolidated text

L 12.7.2024 · 180 articles · 13 annexes

AIO

European Commission · DG CNECT

European AI Office, guidance and codes

Monthly updates · ec.europa.eu

CNIL · French competent authority

CNIL recommendations on the AI Act

Designated Art. 70 · cnil.fr/ia

PDF

aiacto · editorial resource

White paper, read the AI Act in 90 minutes

64-page PDF · v3.1 · May 2026

Frequently asked questions

Everything you need to know about the application schedule.

Concise answers, references to articles of the Regulation. For any specific legal question, consult your counsel or contact your national competent authority.

When did the AI Act enter into force?

Regulation (EU) 2024/1689 entered into force on 1 August 2024, twenty days after its publication in the Official Journal of the European Union (12 July 2024). Entry into force is not equivalent to application: Article 113 provides for successive tiers, now staggered until 2 August 2028 according to the omnibus schedule. See the AI Act glossary for key definitions.

Which obligations are already applicable today?

Under the current schedule, three full tiers are active: prohibited practices and AI literacy since February 2025, GPAI obligations since August 2025, and Article 50 transparency since 2 August 2026. The next major tier covers Annex III high-risk systems on 2 December 2027.

What is the omnibus proposal and how does it affect the schedule?

The omnibus proposal revises the AI Act timetable. The postponement initially considered for transparency did not change the application of Article 50, which has been effective since 2 August 2026. The displayed calendar uses 2 December 2027 for Annex III high-risk systems and 2 August 2028 for Annex I regulated products.

What happens if I am not compliant by the deadline?

Article 99 provides three levels of sanctions: €35M or 7% of worldwide turnover for breach of prohibited practices (Art. 5), €15M or 3% for breach of high-risk obligations (Chapter III) and GPAI, €7.5M or 1% for supply of inaccurate information. Competent authorities may also order market withdrawal.

Do the deadlines apply to AI systems already in service?

Yes, but Article 111 provides transitional provisions. Systems placed on the market before the high-risk deadline are only subject to Chapter III obligations in the event of a subsequent substantial modification. GPAI models placed on the market before 2 August 2025 benefit from an additional period to comply.

How should I prepare for the upcoming deadlines?

The method recommended by the AI Office: (1) map the AI systems deployed or developed, (2) qualify their regulatory role and risk level, (3) build Annex IV documentation for high-risk systems, (4) deploy AI Literacy training. Our free 3-minute diagnostic covers steps 1 and 2.

PREPARE FOR UPCOMING DEADLINES

Identify your obligations in 3 minutes.

Reasoned diagnostic, Annex III classification, exhaustive list of applicable articles, exported as an audit-ready PDF. No sign-up, no credit card.

Start the free diagnostic Talk to an expert

Systems documented

2,400+

Organisations supported

340

Diagnostic duration

3 min

Application scheduleof the European AI Regulation.

Which deadline concerns you?

Everything you need to know about the application schedule.

Identify your obligations in 3 minutes.

Application schedule
of the European AI Regulation.