Reliable Context
Back to Articles|Published on October 28, 2026|15 min read
Download PDF

Reliable Context Article

EU AI Act Dates After the Digital Omnibus (2026/1744)

Which EU AI Act obligations apply when, after Regulation (EU) 2026/1744: high-risk dates, Article 50 marking, GPAI, prohibitions and AI literacy.

By Reliable Context Research

EU AI Act Dates After the Digital Omnibus (2026/1744)

Summary

  1. 01The Digital Omnibus is in force. The general application date remains 2 August 2026, while the core high-risk obligations move to separate dates for Annex III and Annex I systems.
  2. 02Annex III high-risk systems face the core obligations from 2 December 2027; Annex I product systems face them from 2 August 2028.
  3. 03Original prohibitions, AI literacy measures, GPAI obligations and Article 50 already apply. AI literacy changed in substance, while its application date stayed the same.
  4. 04New prohibitions apply from 2 December 2026. Providers of legacy generative systems placed on the market before 2 August 2026 have until that same date to meet Article 50(2) marking requirements.
  5. 05Later high-risk dates do not remove the need to classify systems, start logs and documentation, and record data provenance now. Records that were not kept cannot be produced retroactively.
Inside this article
  1. 01Status as checked on 28 October 2026
  2. 02The original schedule in Article 113
  3. 03What the amended Article 113 says
  4. 04The dated table
  5. 05Prohibited practices and AI literacy
  6. 06General-purpose AI models
  7. 07High-risk systems: Annex III and Annex I
  8. 08Transparency obligations under Article 50
  9. 09What to do now, regardless of dates
  10. 10Checking status: sources to monitor
  11. 11Sources

Status as checked on 28 October 2026

The application dates of the EU Artificial Intelligence Act changed in 2026. The Act itself is Regulation (EU) 2024/1689, which entered into force on 1 August 2024 and was designed to apply in stages. The amending act is Regulation (EU) 2026/1744, the "Digital Omnibus on AI", signed on 8 July 2026 and published in the Official Journal on 24 July 2026. Its final article provides that it enters into force on the third day following publication, and the European Commission confirmed that the AI Omnibus entered into force on 27 July 2026.

The Omnibus is therefore no longer a proposal. The Commission's AI Act policy page traces the path: the proposal was adopted by the Commission on 19 November 2025, a political agreement was reached on 7 May 2026, and the regulation entered into force on 27 July 2026.

This page sets out what applies when, using only the adopted texts and the Commission's own pages. It is an educational summary of EU law as checked on 28 October 2026, not legal advice. Whether a particular system is in scope, and in which risk category, depends on facts this page cannot assess.

Three terms recur:

  • Entry into force is the date a regulation becomes part of the legal order.
  • Date of application is the date from which a provision's obligations bind operators. For the AI Act the two are different, and most of the practical questions are about application.
  • Provider and deployer are the AI Act's two main operator roles: roughly, the party that develops and places an AI system on the market under its own name, and the party that uses it under its authority. The definitions are in Article 3 of the AI Act.
2 August 2026General application date, unchanged by the Omnibus
2 December 2026Application date for the new prohibitions
2 December 2027Core high-risk obligations apply to Annex III systems
2 August 2028Core high-risk obligations apply to Annex I product systems

The original schedule in Article 113

Article 113 is the provision that sets the dates. In the original 2024 text, it said the regulation enters into force on the twentieth day after publication and "shall apply from 2 August 2026", with three exceptions:

  • (a) Chapters I and II (general provisions, including AI literacy, and prohibited practices) from 2 February 2025;
  • (b) Chapter III Section 4 (notifying authorities and notified bodies), Chapter V (general-purpose AI models), Chapter VII (governance), Chapter XII (penalties) and Article 78 (confidentiality) from 2 August 2025, with the exception of Article 101 (fines for providers of general-purpose AI models);
  • (c) Article 6(1) and the corresponding obligations (high-risk systems that are products or safety components of products under the Union harmonisation legislation in Annex I) from 2 August 2027.

Recital 1 of the Omnibus summarises that design: the provisions enter into application in stages, "with all rules entering into application by 2 August 2027". Under the original text, high-risk systems listed in Annex III (biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice and democratic processes) would have been bound from the general date, 2 August 2026.

What the amended Article 113 says

The Omnibus replaced points (a) and (c) of Article 113 and added a point (d). The Commission's AI Act Service Desk shows Article 113 as amended, with each changed point labelled. The amended text reads, in substance:

  • General application: still 2 August 2026.
  • (a) Chapters I and II apply from 2 February 2025, except the new prohibitions in Article 5(1), first subparagraph, points (ba) and (bb), and the new Article 5(1a) and (1b), which apply from 2 December 2026.
  • (b) Unchanged: Chapter III Section 4, Chapters V, VII and XII and Article 78 from 2 August 2025, except Article 101.
  • (c) Chapter III Sections 1, 2 and 3, except Article 6(5), apply from 2 December 2027 for systems classified as high-risk under Article 6(2) and Annex III, and from 2 August 2028 for systems classified as high-risk under Article 6(1) and Annex I.
  • (d) New: Articles 102 to 110 (the amendments the AI Act makes to other Union acts) apply from 27 July 2026.

Recital 40 of the Omnibus gives the reason: delayed availability of standards, common specifications and guidance, and the delayed establishment of national competent authorities, which in the legislator's words "jeopardise the effective entry into application" of the high-risk obligations on the original date.

Chapter III Sections 1 to 3 are the core of the high-risk regime. Section 1 covers classification (Articles 6 and 7). Section 2 holds the requirements: risk management (Article 9), data and data governance (Article 10), technical documentation (Article 11), record-keeping and automatic logs (Article 12), transparency to deployers (Article 13), human oversight (Article 14), and accuracy, robustness and cybersecurity (Article 15). Section 3 holds the obligations of providers, deployers and other parties, including quality management (Article 17), documentation keeping (Article 18), retention of automatically generated logs (Article 19) and deployer obligations (Article 26). Section numbering and article titles are as in the AI Act text.

The dated table

The table lists each obligation group, the original date and the date after the Omnibus. "Unchanged" means the Omnibus did not amend the date, though it may have amended the substance.

Obligation groupMain provisionsOriginal dateDate after OmnibusSource
Entry into force of the AI ActArt. 113, first paragraph1 Aug 2024UnchangedOmnibus, recital 1
General provisions, AI literacy, original prohibitionsChapters I and II (Arts 1 to 5)2 Feb 2025Unchanged (Art. 4 text amended)Art. 113(a) as amended
GPAI model obligations, governance, penalties, notified bodiesChapter III Sec. 4, Chapters V, VII, XII, Art. 782 Aug 2025UnchangedArt. 113(b)
Fines on GPAI model providersArt. 1012 Aug 2026 (general date)UnchangedArt. 113(b)
Amendments to other Union actsArts 102 to 1102 Aug 2026 (general date)27 Jul 2026Art. 113(d), new
General date: transparency, registration, market surveillance, remaining provisionsincl. Chapter IV (Art. 50)2 Aug 2026UnchangedArt. 113, second paragraph
Marking of synthetic content, systems already on the marketArt. 50(2) via new Art. 111(4)2 Aug 20262 Dec 2026Omnibus, Art. 1, point (39)
New prohibitions (non-consensual intimate material, child sexual abuse material)Art. 5(1)(ba), (bb), 5(1a), 5(1b)Did not exist2 Dec 2026Art. 113(a) as amended
National AI regulatory sandbox operationalArt. 57(1)2 Aug 20262 Aug 2027Omnibus, Art. 1, point (22)
GPAI models placed on the market before 2 Aug 2025Art. 111(3)2 Aug 2027UnchangedAI Act, Art. 111(3)
High-risk systems under Annex IIIChapter III Sec. 1 to 32 Aug 20262 Dec 2027Art. 113(c)(i)
High-risk systems under Annex I (products)Chapter III Sec. 1 to 32 Aug 20272 Aug 2028Art. 113(c)(ii)
High-risk systems for public authorities already on the marketArt. 111(2)2 Aug 2030UnchangedOmnibus, Art. 1, point (39)
AI components of large-scale EU IT systems (Annex X)Art. 111(1)31 Dec 2030UnchangedAI Act, Art. 111(1)
Figure 01
Application dates after the Digital Omnibus

  1. 2 Feb 2025Original prohibitions apply

    The eight original prohibitions have applied since this date. The AI literacy application date also remains unchanged.

  2. 2 Aug 2025GPAI model obligations apply

    Chapter V obligations apply to GPAI model providers. Models placed on the market before this date have until 2 August 2027.

  3. 2 Aug 2026General application date

    The general date remains unchanged. Article 50 transparency obligations apply from this date.

  4. 2 Dec 2026New prohibitions and legacy marking

    New prohibitions apply. Providers of generative systems placed on the market before 2 August 2026 must meet Article 50(2) marking requirements.

  5. 2 Dec 2027Annex III high-risk obligations

    Chapter III Sections 1 to 3 bind stand-alone high-risk uses classified through Article 6(2), sixteen months after the original date.

  6. 2 Aug 2028Annex I high-risk obligations

    The core high-risk obligations bind AI that is a regulated product or a safety component of one, twelve months after the original date.

Later application dates do not change what evidence a high-risk system will need to produce on the day it applies. Most of that evidence is a record of the past: the data a model was trained and validated on, the events a system logged, the documentation that existed when a decision was made. Records that were not kept cannot be produced retroactively.

Prohibited practices and AI literacy

The eight original prohibitions in Article 5 (for example, certain manipulative techniques, social scoring and untargeted scraping of facial images) have applied since 2 February 2025. The Commission's AI Act page lists them and notes that a ninth prohibition "comes into effect in December 2026 and was introduced as a part of the AI Omnibus".

The new points in the Omnibus prohibit placing on the market, putting into service or using an AI system that generates or manipulates realistic intimate images, video or audio of an identifiable person without that person's explicit consent (point (ba)), or that generates child sexual abuse material within the meaning of Directive 2011/93/EU (point (bb)). New paragraph 1a limits the prohibition on placing such systems on the market to cases where that generation is the intended purpose or a reasonably foreseeable and reproducible outcome without adequate safeguards. These apply from 2 December 2026.

AI literacy changed in substance, not in date. The original Article 4 required providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff. The replacement text in the Omnibus requires them to "take measures to support the development of AI literacy" of staff and others operating AI systems on their behalf, and adds that the obligation "does not require providers or deployers to guarantee any specific level of AI literacy of any individual". The Commission and Member States are tasked with supporting the effort, including by publishing practical examples.

General-purpose AI models

The obligations on providers of general-purpose AI (GPAI) models in Chapter V have applied since 2 August 2025, and the Omnibus did not move that date. Models placed on the market before 2 August 2025 have until 2 August 2027 under Article 111(3). The Commission's power to fine GPAI model providers under Article 101 was carved out of the 2025 date and therefore follows the general date of 2 August 2026, per Article 113(b).

The General-Purpose AI Code of Practice was published on 10 July 2025. The Commission describes it as a voluntary tool with three chapters (transparency, copyright, and safety and security) and states that the Commission and the AI Board have confirmed it is an adequate voluntary tool for providers to demonstrate compliance. The Commission's AI Act page adds that from 2 August 2026 the AI Office holds enforcement powers over GPAI models, including requesting technical documentation, evaluating models and issuing fines.

High-risk systems: Annex III and Annex I

The two-track delay is the change with the largest planning consequence. Under the amended Article 113(c):

  • Annex III systems (stand-alone high-risk uses, classified through Article 6(2)) are bound by Chapter III Sections 1 to 3 from 2 December 2027, sixteen months after the original date.
  • Annex I systems (AI that is itself a regulated product or a safety component of one, classified through Article 6(1)) are bound from 2 August 2028, twelve months after the original date.

The Commission's news item on entry into force summarises the same two dates, with examples: biometrics, critical infrastructure, education, employment, migration, asylum and border control for the first, and machinery, toys and lifts for the second.

Two transitional rules matter for systems already in use. First, the amended Article 111(2) now ties the grandfathering of existing high-risk systems to "the date of application of Chapter III referred to in Article 113" rather than to the fixed 2 August 2026: systems placed on the market or put into service before that date are caught only if their design changes significantly afterwards. Recital 39 clarifies that the grace period attaches to the type and model: if one unit of a type was lawfully placed on the market before the date, further identical units benefit too, as long as the design stays unchanged. Second, high-risk systems intended for use by public authorities must still comply by 2 August 2030 regardless.

The Omnibus also amended the substance of some Section 2 requirements. For example, it replaced Article 10(1) and deleted Article 10(5), moving the exceptional processing of special categories of personal data for bias detection into a new Article 4a [1]. A team that drafted against the 2024 text should re-read the amended articles, not only the dates.

Registration is the provision most often misread. The Omnibus did not amend Article 49 or move its date, and Article 6(4) still makes a provider that considers an Annex III system not high-risk subject to registration under Article 49(2). What changed is the content: points 7 and 9 of Annex VIII, Section B, are deleted, which recital 22 presents as streamlining that registration [1]. Article 49 sits in Chapter VIII, which keeps the general date of 2 August 2026. The text does not say how the registration of Annex III high-risk systems operates in the period before their substantive obligations apply on 2 December 2027, and no Commission guidance on that point was found when this page was checked on 28 October 2026.

A worked example shows the arithmetic. Halden Therapeutics AG, a fictional pharmaceutical company used for illustration, deploys three AI systems in the EU:

System (fictional)Likely route to high-risk statusEarliest date Chapter III Sec. 1 to 3 bind it
CV-screening model used to shortlist job applicantsAnnex III, point 4 (employment)2 December 2027
AI module that is a safety component of a medical device subject to third-party conformity assessmentArticle 6(1), Annex I (medical devices regulation)2 August 2028
Chat assistant that answers staff questions about internal SOPsNot high-risk on these facts; Article 50(1) disclosure may applyArticle 50 from 2 August 2026

The classification column is illustrative. Whether a given system is high-risk depends on Article 6 and the Commission's guidelines, not on a label.

Transparency obligations under Article 50

Article 50 sits in Chapter IV and was not moved: it applies from the general date. The Commission's guidelines page on transparency obligations states that Article 50 "applies from 2 August 2026" and covers providers and deployers of certain AI systems, including generative and interactive systems and deepfakes.

The Omnibus added one transitional rule. New Article 111(4) gives providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text and that were placed on the market before 2 August 2026, until 2 December 2026 to comply with Article 50(2), the obligation to mark outputs in a machine-readable format and make them detectable as artificially generated. Recital 38 describes this as a four-month transitional period.

The Omnibus also rewrote Article 50(7): the Commission encourages codes of practice on detecting, marking and labelling generated content, assesses whether adherence to them is adequate, and may adopt an implementing act with common rules if a code is inadequate. The Commission's guidelines page links a Code of Practice on Transparency of AI-generated Content and states that providers and deployers who do not adhere to it will have to demonstrate compliance by other means.

What to do now, regardless of dates

Later application dates do not change what evidence a high-risk system will need to produce on the day it applies. Most of that evidence is a record of the past: the data a model was trained and validated on, the events a system logged, the documentation that existed when a decision was made. Records that were not kept cannot be produced retroactively. Four practical steps follow from the texts above:

  1. Classify now and date the classification. Record which systems you treat as Annex III, Annex I, Article 50 only, or out of scope, with the reasoning and the version of the text used. The answer may change with guidance; the record of what you decided and when should not.
  2. Start logs and documentation before the obligation date. Article 12 requires high-risk systems to technically allow automatic recording of events over their lifetime, and Article 19 requires providers to keep those logs [2]. An audit trail that begins on 2 December 2027 cannot show how a system behaved before it.
  3. Record the provenance of training and reference data. Article 10 data-governance requirements ask about origin, collection and preparation of data sets. The provenance of each data set (source, date obtained, transformations, responsible party) is far cheaper to capture at ingestion than to reconstruct later.
  4. Handle the dates that already passed. Prohibitions, AI literacy measures, GPAI obligations and Article 50 already apply. The 2 December 2026 date for the new prohibitions and for Article 50(2) marking of legacy generative systems is weeks away at the time of writing.
Figure 02
Prepare evidence before the obligation date
  1. 01Classify and date the decision

    Record each system's category, your reasoning and the version of the text used. Keep the record of what you decided and when.

  2. 02Start logs and documentation

    Start before the obligation date so the audit trail can show how the system behaved earlier.

  3. 03Record data provenance

    Capture each data set's source, date obtained, transformations and responsible party at ingestion.

  4. 04Handle dates already passed

    Address the prohibitions, AI literacy measures, GPAI obligations and Article 50 requirements that already apply.

Checking status: sources to monitor

Dates in this field have changed once and can change again only by another amending act. The sources that would show a change first:

A summary like this one is a secondary source. When a date drives a budget or a contract, read the article in the Official Journal text and note the date you read it. Further reading on the site: frequently asked questions.

Reliable Context is a system of record in which every value carries its cited, dated source, which is the kind of evidence these provisions ask organizations to keep.

Sources

  1. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (Artificial Intelligence Act), OJ L, 2024/1689, 12.7.2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  2. Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 (Digital Omnibus on AI), OJ L, 2026/1744, 24.7.2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj
  3. European Commission, AI Act Service Desk, Article 113: Entry into force and application (consolidated display with Omnibus amendments marked). https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-113
  4. European Commission, "AI Omnibus enters into force", news article, 27 July 2026. https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force
  5. European Commission, "AI Act" (regulatory framework for AI), policy page, accessed 28 October 2026. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
  6. European Commission, "The General-Purpose AI Code of Practice", policy page, accessed 28 October 2026. https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai
  7. European Commission, "Guidelines on transparency obligations for providers and deployers of certain AI systems", policy page, 6 August 2026. https://digital-strategy.ec.europa.eu/en/policies/guidelines-transparency-ai-generated-content
External Sources (2)
[1]eur-lex.europa.eu/eli/reg/2026/1744/oj
[2]eur-lex.europa.eu/eli/reg/2024/1689/oj
eu ai act timelineai act application datesdigital omnibus ai act delayai act high risk deadlineai act 2 august 2026

Published by

Reliable Context Research

Reliable Context is a system of record for organizations that fuse private and public information and must defend every value they act on: one canonical record per real thing, ambiguity shown rather than hidden, and provenance behind every value.

Our product · Evidence and editorial standards · Contact

Request access

Editorial note

Researched and written by Reliable Context Research with AI-assisted research tools; sources are cited inline and checked before publication. Report an error: https://reliablecontext.com/corrections.