AI disclosure and Article 50
When the target-language lead vocal is produced from the artist’s authorised voice, the release needs the right label and, where the law requires it, a human-readable disclosure at first exposure. This guide explains the IFPI labels, the EU AI Act Article 50 duty and what Moozz supplies with a release package.
Last reviewed by Moozz on 2026-10-21
This is general information, not legal advice.
What the release needs
- The IFPI label for machine-made recordings where the release uses the IFPI programme.
- A human-readable disclosure at first exposure where Article 50 or a platform requires it.
- The disclosure in the release metadata and credits.
- The artist voice consent and the core permission record behind the version.
IFPI labels for machine-made and machine-assisted recordings
IFPI’s voluntary labels describe how the recording was made. The target-language lead vocal in a Moozz version is produced from the artist’s authorised voice. The artist does not perform that vocal. That puts the release in the machine-made category, not the machine-assisted category.
| Label | What it means | When it applies |
|---|---|---|
| Machine-made recording | Machine systems produced all or the primary portion of the creative elements, including the lead vocal. | The target-language lead vocal is produced by machine. |
| Machine-assisted recording | Humans performed the lead vocal and primary instruments; machines helped with some expressive elements. | A human sings the target-language lead vocal. |
Article 50 in plain language
Article 50 of the EU AI Act has two relevant duties. Providers of generative AI systems must mark synthetic audio in a machine-readable format. Deployers that generate or manipulate audio that looks or sounds authentic must disclose that it is machine-made.
The artistic-work exception limits how the disclosure is made; it does not remove the duty. A clear label at first exposure, plus the credit and metadata, is the practical route. Article 50 is a disclosure duty, not a certification, and it does not replace the rights permissions.
What Moozz supplies
- The machine-made designation in the release metadata where the IFPI programme is used.
- A plain-language disclosure for first exposure where the release is in scope.
- The release package and the consent record behind the version.
- A link to Liner, the free Article 50 disclosure tool, for teams that need to document the release.
What the label or distributor still does
The releasing label or distributor displays the disclosure at first exposure and preserves the metadata through delivery. Moozz supplies the wording and the record; the release channel decides where the listener sees it.
If you are not sure
Article 50 is conditional and territorial. Ask your counsel about the release territories and the platform requirements. This guide is general information, not legal advice.
Sources
Questions and answers
Is the target-language lead vocal machine-made or machine-assisted?
It is machine-made. The target-language lead vocal is produced from the artist’s authorised voice; the artist does not perform that vocal. Machine-assisted would require a human to perform the lead vocal.
Does consent remove the need for disclosure?
No. Consent makes the use lawful if the rights are cleared. It does not remove the Article 50 disclosure duty where the audio is in scope.
What is Liner?
Liner is a free tool that writes the Article 50 transparency disclosure a release needs. It is open to anyone, including teams that do not use Moozz.
Does Moozz guarantee compliance?
No. Moozz supplies the label, the disclosure where required and the rights record. The release channel and its counsel remain responsible for the final release.
Release your song in a new language
Send us your master recording and the target language. We will confirm the rights and consents and scope the work.