Quick answer: The Code is voluntary. Article 50 is not. The European Commission published the Code of Practice on Transparency of AI-Generated Content on 10 June 2026. The Code is voluntary implementation support for Article 50 marking and labelling workflows. Article 50 transparency obligations remain legal obligations where they apply, with the Commission listing 2 August 2026 as the application date for those obligations.
What changed on 10 June 2026
The Code moved from a consultation workstream to a Commission-published voluntary Code of Practice. That matters because deployers now have a stronger practical reference for labelling deepfakes and certain AI-generated or manipulated public-interest text. Providers have a clearer reference for marking and detection of AI-generated or manipulated outputs.
The legal baseline did not become voluntary. Article 50 remains the source of the transparency obligations. The Code can support implementation and evidence, but it does not replace Regulation (EU) 2024/1689, Commission Article 50 guidelines, or qualified legal review.
Article 50 law vs the Code of Practice
| Question | Article 50 law | Code of Practice |
|---|---|---|
| Legal status | Binding legal obligations where Article 50 applies. | Voluntary implementation support. |
| Main function | Defines transparency triggers for providers and deployers. | Gives practical marking and labelling measures for signatories and implementation planning. |
| Provider focus | Machine-readable marking and detectability for AI-generated or manipulated outputs. | Section 1: provider marking and detection rules. |
| Deployer focus | Disclosure 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 consequence | Retain scenario, trigger, notice, timing, accessibility, owner, and review evidence. | Record Code measure considered, EU icon use, provider dependency, label text, and exception rationale. |
What the Code covers
The Code has two sections. Section 1 addresses providers and rules for marking and detection of AI-generated and manipulated content. Section 2 addresses deployers and rules for labelling deepfakes and AI-generated or manipulated text. The Commission also provides EU icons that deployers may use when labelling AI-generated content.
What deployers should do now
Start with a scenario record. Identify whether the AI use case involves direct AI interaction, synthetic content, deepfake content, public-interest text, emotion recognition, or biometric categorisation. Then record the label or notice wording, placement, first-interaction or first-exposure timing, accessibility check, provider marking dependency, EU icon decision, exception rationale, approval owner, and review date.
For operational routing, use the Article 50 disclosure decision tree, the Article 50 disclosure checklist, and the Article 50 Code vs law comparison.
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
Use these questions to separate voluntary Code measures from Article 50 legal obligations and deployer evidence records.
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.
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.
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.
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.