ARGOArticle 5030 seconds
Who has to label an AI video?
The tool marks the file so a machine can recognise it. You label the deepfake. These are two different obligations.
In short
Two obligations usually travel mixed together. The episode separates them in thirty seconds.
The first one sits with the generating tool: mark the file so a machine can recognise it. The second sits with whoever publishes a deepfake: disclose it.
What you'll learn
- Machine marking is the generator's duty, not the user's.
- Visible disclosure concerns deepfakes, not every image produced by an AI.
- Mixing the two leads either to over-labelling or to labelling nothing at all.
The steps
Step 1 / 2
What the tool does
The tool that generates the video has to mark the file so a machine can recognise its origin.
That marking is not meant to be read by the viewer: it is a technical trace.
Step 2 / 2
What the publisher does
Publishing a deepfake means disclosing that it comes from an AI. The obligation covers that precise case.
It does not extend to every image produced by an artificial intelligence.
What went wrong
- The episode explains the rule without showing the legal text on screen; the repository keeps no capture of the cited source.
- The original soundtrack uses sounds whose rights are unverified and should be replaced before any commercial use.
Evidence in the repository
The files that document this episode. They are not public, but they exist and they carry the numbers quoted here.
Le relevé de publication de l'épisode, validé le 12 août 2026
runs/pilot-002/release.json
Behind the scenes
This episode came out of a pipeline, not twelve browser tabs.
Research, script, shots, media, editing, checks and publishing kit: ARGO Studio holds everything that produced this video.