Transparency (Art. 50)
Since 2 August 2026, Art. 50 sets transparency duties for AI that talks with people or makes content. They apply whatever the risk class of the system.
The duties, case by case
| Case | Who | What |
|---|---|---|
| AI that interacts with people, such as a chatbot | Provider | Design it so people are told they are interacting with AI, unless it is obvious (Art. 50(1)) |
| AI that generates audio, images, video or text | Provider | Mark the output in a machine-readable way as artificially generated (Art. 50(2)) |
| Emotion recognition or biometric categorisation | Deployer | Inform the people exposed to it (Art. 50(3)) |
| Deepfakes | Deployer | Disclose that the content was artificially generated or manipulated (Art. 50(4)) |
| AI-generated text published to inform the public on matters of public interest | Deployer | Disclose it, unless a person reviewed it and someone holds editorial responsibility (Art. 50(4)) |
Older systems: marking by 2 December 2026
AI systems placed on the market before 2 August 2026 must comply with the marking duty of Art. 50(2) by 2 December 2026 (Art. 111(4) as amended).
Guidance
The Commission published guidelines on Art. 50 on 22 July 2026. A voluntary code of practice on marking and labelling AI-generated content dates from 2 July 2026.
This site
The pages here are generated by AI and say so on every page, with the date they were checked against their sources.
Art. 50 label checklist
| Question | Yes / No | Where you tell people |
|---|---|---|
| Do people chat or talk with an AI system of ours? | ||
| Do we publish images, audio or video made or changed by AI? | ||
| Do we publish AI-generated text on matters of public interest without human editorial review? | ||
| Do we use emotion recognition or biometric categorisation? | ||
| Does a tool we provide generate content that must be machine-marked? |
Open the Art. 50 checklist
The same checklist with version and check date, ready to print.