EU AI Act: Deadlines 2026 & 2027 (Which Obligations Apply When)

By Alexander Weipprecht · As of July 2026

The AI Act (Regulation (EU) 2024/1689) does not apply from a single date but in stages. It has been in force since August 1, 2024, yet its obligations take effect under Article 113 in four stages: prohibited practices and AI literacy since February 2025, GPAI and governance since August 2025, and most high-risk obligations from 2026 and 2027.

When Does the AI Act (EU AI Act) Apply?

The AI Act has been in force since August 1, 2024, as the 20th day after publication in the Official Journal on July 12, 2024 (Art. 113 EU AI Act). However, entry into force is not the day on which all obligations take effect. Article 113 sets out a staggered application: the general date is August 2, 2026 ("It shall apply from 2 August 2026"), while individual chapters apply earlier or later. The following table gives an overview of the applicable timeline.

Date	What applies	Legal basis
August 1, 2024	Regulation enters into force	Art. 113
February 2, 2025	Prohibited AI practices + AI literacy	Art. 113 (Ch. I & II)
August 2, 2025	GPAI models, governance, sanctions	Art. 113 (Ch. V, VII, XII; Art. 78)
August 2, 2026	General application date + high-risk Annex III	Art. 113 in conjunction with Art. 6(2)
August 2, 2027	High-risk Annex I (AI in regulated products)	Art. 113 in conjunction with Art. 6(1)

What Is the Difference Between "Entry into Force" and "Application Date"?

Entry into force and application date are two different dates in EU law, and this is exactly why not everything applies from day 1. The AI Act has been legally in existence and bindingly enacted since August 1, 2024 (Art. 113). The application date determines when the individual obligations are actually enforced. The EU legislature deliberately separated these dates so that providers, deployers, and authorities have time to prepare for the requirements. In practice this means: the clock has been running since August 2024, but the concrete behavioral obligations are tied to the staggered application dates in Art. 113.

Which Obligations Have Applied Since February 2, 2025?

Since February 2, 2025, the prohibited AI practices under Art. 5 and the obligation to ensure AI literacy under Art. 4 have applied. Art. 113 states: "Chapters I and II shall apply from 2 February 2025" (Art. 113). Chapter II (Art. 5) prohibits certain AI practices such as manipulative social scoring or impermissible biometric surveillance. Chapter I (Art. 4) requires providers and deployers to ensure a sufficient level of AI literacy among their staff.

Important in practice: the AI literacy obligation does not require a certificate. The European Commission states explicitly: "There is no need for a certificate. Organisations can keep an internal record of trainings," and there is no prescribed form, "no one size fits all" (European Commission, AI literacy Q&A). AI literacy is defined in Art. 3(56) as the skills and knowledge that enable providers, deployers, and affected persons to make informed use of AI and to understand its opportunities and risks. How deep a company goes remains an employer decision; a structured AI training program with verifiable proof of competence is one workable way to document the obligation, though it is not legally required.

What Has Applied Since August 2, 2025?

Since August 2, 2025, the rules for general-purpose AI models (GPAI), the governance structure, and the sanctions framework have applied. Specifically, Art. 113 states that Chapter III, Section 4, Chapter V (GPAI models), Chapter VII (governance), Chapter XII (sanctions), and Art. 78 have applied since this date (Art. 113). One exception is Art. 101: fines for GPAI providers are excluded from the application date of August 2, 2025. This second stage builds the institutional foundation, such as the AI Office and the national competent authorities, before the broad obligations take effect from 2026.

What Changes on August 2, 2026?

On August 2, 2026, the general application date of the Regulation takes effect, and, under the regulatory text currently in force, so do the obligations for standalone high-risk systems under Annex III. Under the political Omnibus agreement of May 7, 2026, however, these Annex III obligations are expected to take effect only on December 2, 2027; until this change formally enters into force, the previous date of August 2, 2026 continues to apply (as of July 2026). These systems fall under the general date of August 2, 2026 via Art. 6(2), not under the extended Annex I deadline (European Commission, Regulatory framework for AI). Annex III lists eight areas of high-risk AI: biometrics; critical infrastructure; general and vocational education; employment and personnel management; access to essential private and public services (including creditworthiness); law enforcement; migration and border control; and the administration of justice and democratic processes (Annex III EU AI Act). Anyone offering or operating such a system must meet the high-risk requirements from this date, such as risk management, data quality, and human oversight.

What Applies Only From August 2, 2027?

Under the regulatory text currently in force, the obligations for high-risk AI under Annex I, meaning AI as a safety component or in products already regulated, apply from August 2, 2027; under the political Omnibus agreement of May 7, 2026, they are expected to be postponed to August 2, 2028, as long as this change has not formally entered into force (as of July 2026). Art. 113 explicitly extends the deadline here: "Article 6(1) … shall apply from 2 August 2027" (Art. 113). Annex I systems are AI embedded in products that are already subject to EU product-safety harmonization today (for example machinery or medical devices). For this category, the legislature grants an additional year, because the AI requirements must be interlinked with existing conformity procedures.

What Fines Are Possible, and When Do 35 Million EUR or 7% Apply?

The highest fine bracket, up to 35,000,000 EUR or 7% of worldwide annual turnover (whichever is higher), applies exclusively to violations of the ban on AI practices under Art. 5, as set out in Art. 99(3) (Art. 99 EU AI Act). This bracket therefore does not apply across the board to every violation, and in particular not to the AI literacy obligation under Art. 4. Other violations are subject to lower tiers.

Fine tier	Applies to
up to 35 million EUR / 7%	Prohibited practices (Art. 5)
up to 15 million EUR / 3%	Other obligations (including Art. 16, 26, 50)
up to 7.5 million EUR / 1%	False or misleading statements to authorities

For SMEs and start-ups, the lower of the two values applies (the amount or the percentage), a deliberate relief measure for smaller companies (Art. 99).

Does the "Digital Omnibus" Postpone the High-Risk Deadlines for 2026/2027?

As of June 9, 2026, the applicable Art. 113 deadlines (August 2, 2026 and August 2, 2027) continue to apply unchanged; a postponement has been proposed but has not yet entered into force. The "Digital Omnibus on AI" (Commission proposal of November 19, 2025) provides for postponing standalone Annex III systems to December 2, 2027 and Annex I systems embedded in products to August 2, 2028 (Gibson Dunn, EU AI Act Omnibus). In early May 2026, the Council and the Parliament announced a preliminary political agreement to simplify the rules (Council of the EU). However, this is so far only a preliminary agreement; formal adoption and publication in the Official Journal are, according to the sources cited here, still pending (White & Case). Until then, the applicable deadlines remain authoritative.

FAQ: Common Questions About the AI Act Deadlines

Since When Has the EU AI Act Been in Force?

The AI Act (EU) 2024/1689 has been in force since August 1, 2024, as the 20th day after publication in the Official Journal on July 12, 2024 (Art. 113).

From When Does the EU AI Act Generally Apply?

The general application date is August 2, 2026. Individual chapters apply earlier (since February 2, 2025 and August 2, 2025) or later (August 2, 2027), as staggered by Art. 113.

Since When Has the AI Literacy Obligation Applied?

The AI literacy obligation under Art. 4 has applied since February 2, 2025. It does not require a certificate; an internal record of training is sufficient according to the European Commission.

Do I Need a Certificate for AI Literacy?

No. The European Commission makes clear: "There is no need for a certificate." There is no prescribed form; the appropriate depth remains the employer's decision.

When Do the High-Risk Obligations Apply?

Standalone high-risk systems under Annex III (Art. 6(2)) from August 2, 2026; systems embedded in regulated products under Annex I (Art. 6(1)) from August 2, 2027.

When Do Fines of 35 Million EUR or 7% Apply?

This highest bracket applies only to violations of the prohibited AI practices under Art. 5 (Art. 99(3)). Other violations are sanctioned with lower tiers (15 million EUR/3% or 7.5 million EUR/1%).

Have the 2026/2027 Deadlines Already Been Postponed?

No. As of June 2026, there is only a preliminary agreement on the "Digital Omnibus"; the applicable deadlines of August 2, 2026 and August 2, 2027 continue to apply until formal adoption.

Sources

• EU AI Act, Art. 113: Entry into Force and Application Date (artificialintelligenceact.eu, 2024)
• EU AI Act, Art. 99: Fines (artificialintelligenceact.eu, 2024)
• EU AI Act, Annex III: High-Risk Areas (artificialintelligenceact.eu, 2024)
• European Commission: AI literacy, Questions and Answers (2025)
• European Commission: Regulatory framework for AI
• Council of the EU: Press release on simplifying the AI rules (May 2026)
• White & Case: EU agrees Digital Omnibus deal (2026)
• Gibson Dunn: EU AI Act Omnibus Agreement (2025)

As of July 2026. This article provides general information and does not replace legal advice. The applicable regulatory text (EUR-Lex) is authoritative in each case.