AI Music Policy

1. What we will not distribute

Two categories of submissions are rejected outright. Both apply regardless of plan, label, or prior catalogue with wavbee:

• Tracks that are entirely AI-generated — including audio produced end-to-end by tools like Suno, Udio, or comparable platforms with no human composition, performance, or production input
• Tracks built on AI models trained on copyrighted material without the rights holders' permission, regardless of how much human work was layered on top afterwards

We do not litigate the line case-by-case. If your release would not exist without a single AI generation step, it is not accepted here.

2. AI as a tool, not the artist

AI-assisted music is welcome on wavbee when AI is one of several tools you used, not the track itself. Examples we accept:

• AI-assisted mixing, mastering, or stem separation (e.g., LANDR, iZotope, Lalal.ai)
• AI-generated MIDI or stem ideas that you then performed, edited, arranged, and produced
• AI-assisted vocal tuning or harmonization on a recording you sang
• AI-generated stems or samples cleared for commercial use, integrated into your composition

The simple test: if you removed the AI step, would there still be a track? If yes, you are fine.

3. Models must be fairly trained

When AI is involved at any stage, the underlying model must be trained on data the model operator has the legal right to use. The wider music industry has aligned around this principle and so do we:

• Models trained on licensed datasets, public-domain audio, or the operator's own catalogue are fine
• Models trained by scraping copyrighted recordings from streaming services or the open web are not
• The Fairly Trained certification (fairlytrained.org) is a useful signal but is not strictly required — any model whose training corpus is licensed or public-domain qualifies
• If you cannot verify how your AI tool was trained, do not assume it is fine — ask the vendor or use a different tool

The legal risk of distributing a track built on a model trained on unlicensed data is yours, not ours. We will remove the track and apply our standard violation procedure if a rights holder credibly demonstrates infringement.

4. Disclosure at submission

Every release form asks whether AI was used in any part of the track. Answer truthfully:

• Disclose AI-assisted mixing, mastering, stem separation, MIDI ideation, or vocal processing
• Disclose the specific tools and which steps each one contributed to
• Disclose the licensing position of any AI-generated material you included
• Undisclosed AI use is treated the same as undisclosed sampling — see the Terms of Service section on fraud and misrepresentation

5. Enforcement and consequences

We act on this policy proactively and retroactively:

• New submissions that disclose ineligible AI use are declined at the application stage and the verification charge is not refunded
• Submissions that omit disclosure and are later found to involve ineligible AI use are taken down from all DSPs
• Repeated or fraudulent submissions result in permanent account termination and royalty forfeiture, per the Terms of Service
• We follow industry guidance from the major DSPs — their policies evolve faster than ours, and we may tighten this policy without notice when they do

If you are unsure whether a specific tool or workflow is acceptable, ask us at [email protected] before you submit. We would rather have a quick conversation than reject a release after the fact.