EU AI Act Article 50 timeline: code of 10 June, duties from 2 August
Article 50 transparency duties apply from 2 August 2026. The final labelling code was published 10 June 2026, with about 190 signatories by end of July.

The Commission's page says the Article 50 transparency obligations apply from 2 August 2026. The final Code of Practice on marking and labelling AI-generated content was published on 10 June 2026, after drafts on 17 December 2025 and 3 March 2026, and about 190 organisations had signed by the end of July 2026. For anyone publishing AI video in the EU, the start date is already behind us.
The dates in order
The page is the Commission's own summary. Other dates on related pages are not part of this table.
| Date | Event |
|---|---|
| 17 December 2025 | First draft code published |
| 3 March 2026 | Second draft published |
| 10 June 2026 | Final code published |
| 2 August 2026 | Article 50 transparency obligations apply |
| End of July 2026 | About 190 organisations signed |
Reading the timeline as a producer
Three things follow for a team that makes generated video. First, the code is final, so you can read it rather than waiting for it. Second, the legal duties are live, so an assessment of your own role is overdue. Third, a large signatory list means many platforms and tools are working to the same code.
Your role depends on what you do. Running a generator is different from publishing its output, and the code treats those differently. The video docs describe generating a clip through the API; what you do with the file afterwards is the publishing side.
A short checklist
This is a summary of a Commission page, not legal advice. For an actual decision, read the code and the regulation text, or ask counsel.
- Write down which of your outputs are realistic synthetic video, audio or images.
- Mark who in your team publishes them and where.
- Read the final code's sections for providers and for deployers.
- Keep a dated record of each clip: job, prompt, date, label applied.
Using the dates
The sequence shows how the work moved: two drafts, a final code, then the application date. Teams that started reading the drafts in December had months to adjust; teams starting now are late but not blocked.
Start with an inventory. List the AI-generated content you publish, the channel, the audience country and who approves it. Then map each row to a duty. Most video teams will find a short list of rows with real obligations.
Review the inventory on a schedule, such as quarterly, and whenever you add a new content type. A generated voice or a new avatar format changes your rows.
Sources
Related posts
More in Use cases
- Event RSVP clips: invitation, reminder and last call from one avatar
Plan three Sume avatar clips for one event, an 18-second invitation, a 10-second reminder and an 8-second last call. 36 seconds total costs $6.62 on Standard.
- Reels ads: 9:16, sound on, and a safe-zone brief for generated clips
Meta says Reels default to sound on and favours 9:16 video with audio and key messages in the safe zone. A brief for generating clips that follow it.
- Faire product video: three clips, 2 GB, 1080p
Faire's video rules: 1920x1080 or larger, up to 2 GB, up to three videos. How to batch 16:9 product clips on Sume and check each file.
- Food bank volunteer call for Thanksgiving: a 15-second clip, $0.41
Make a silent, captionable volunteer call for a turkey drive before Thursday, November 26, 2026: three stills, a music bed and one Timeline render on Sume.
Written by Sume