EU AI Act update: the Council adopted the Digital Omnibus on AI on 29 June 2026. Official Journal publication is still pending; review the official Council update before relying on older high-risk deadline references. Council update EU AI Act update: Council adopted the Digital Omnibus; OJ publication pending. Council update

Article 50 legal-status clarification

Article 50 Code vs Article 50 Law: What Is Legal Obligation and What Is Implementation Support

Regulatory Engine Active: Evaluated against EU AI Act ruleset v3.8. Rules engine updated: June 23, 2026. Full statutory baseline last reviewed: June 15, 2026.
Current-law EU AI Act dates are separated from Digital Omnibus planning material, the final voluntary Article 50 Code of Practice, signatory and non signatory evidence routes, EU icon use, and draft high-risk classification guidance. The Code can support marking and labelling implementation, but it does not replace in-force law, Commission guidelines, or legal review.

Quick answer: The Code is voluntary. Article 50 is not. Article 50 is binding law where it applies. The Article 50 Code of Practice is voluntary implementation support published by the European Commission on 10 June 2026. The Code helps structure marking, detection, and labelling workflows, but it does not make Article 50 optional and does not replace legal review.

Article 50 law vs the Code of Practice

QuestionArticle 50 lawCode of Practice
Legal statusBinding legal obligations where Article 50 applies.Voluntary implementation support.
Main functionDefines transparency triggers for providers and deployers.Gives practical marking and labelling measures for signatories and implementation planning.
Provider focusMachine-readable marking and detectability for AI-generated or manipulated outputs.Section 1: provider marking and detection rules.
Deployer focusDisclosure for deepfakes and certain AI-generated or manipulated public-interest text, plus other Article 50 notice duties.Section 2: deployer labelling of deepfakes and AI-generated or manipulated text.
Evidence consequenceRetain scenario, trigger, notice, timing, accessibility, owner, and review evidence.Record Code measure considered, EU icon use, provider dependency, label text, and exception rationale.

Provider marking vs deployer labelling

Provider marking is mainly about machine-readable marking and detectability of AI-generated or manipulated outputs. Deployer labelling is about clear disclosure to people when deepfakes or certain AI-generated or manipulated public-interest text are published. Many organisations will need both vendor evidence and their own public-facing notice record.

The deployer mistake to avoid

Do not assume that a provider watermark, provenance signal, or EU icon fully handles deployer obligations. The deployer still needs to classify the scenario, decide whether a notice or label is required, approve wording, confirm timing and accessibility, and retain evidence of the decision.

Article 50 evidence fields to add after the Code update

  • Article 50 scenario: direct AI interaction, synthetic content, emotion recognition, biometric categorisation, deepfake, or public-interest AI-generated/manipulated text.
  • Code measure considered and whether the organisation plans to sign or follow a Code-aligned route.
  • Provider marking dependency and provider evidence requested.
  • EU icon considered or used, with a note that icon use alone is not legal proof.
  • Human-facing label text, label placement, first-interaction or first-exposure timing, and accessibility check.
  • Public-interest text review outcome, human review or editorial responsibility, exception rationale, approval owner, date, and screenshot or record location.

Article 50 Code update FAQ

Is the Article 50 Code of Practice mandatory?

No. The Code is voluntary implementation support. Article 50 transparency obligations remain legal obligations where they apply. Use the Code to structure marking, labelling, evidence, and review workflows. Do not treat Code participation as a substitute for the AI Act, Commission guidelines, or legal review.

What changed on 10 June 2026?

The Commission published the Code of Practice on Transparency of AI-Generated Content. The practical change for deployers is not a new standalone law. It is a clearer implementation reference for labelling deepfakes and certain AI-generated or manipulated public-interest text, plus evidence fields such as label wording, placement, icon consideration, and accessibility.

What is the difference between provider marking and deployer labelling?

Provider marking concerns machine-readable marking and detection for AI-generated or manipulated outputs. Deployer labelling concerns public-facing disclosure for deepfakes and certain AI-generated or manipulated text published to inform the public on matters of public interest. A deployer evidence file should show which side of that split was reviewed.

Do EU icons prove Article 50 compliance?

No. EU icons can support clearer labelling, but the deployer still needs a disclosure process. Record where the label appears, when the person sees it, whether it is accessible, what text accompanies it, and who approved the notice. Icon use alone should not be described as compliance proof.

Source and review note

Last reviewed: 15 June 2026. Source basis: Regulation (EU) 2024/1689 Article 50, the European Commission Code of Practice on Transparency of AI-Generated Content published on 10 June 2026, European Commission signing instructions for the Code, and the European Commission EU icons page for labelling AI-generated content. The Code is voluntary implementation support. Article 50 transparency obligations remain legal obligations where they apply. EU icons are optional labelling aids and do not establish legal compliance by themselves. This page provides operational planning guidance, not legal advice, certification advice, audit assurance, or a compliance guarantee.