Guides

AI disclosure in 2026: what to fill in, and what a mastering tool should never do

Updated 2026-09-29

The rules changed during 2026. Here is the state of play, in plain terms, and where mastering fits.

What distributors ask now

DistroKid, TuneCore and most others include an AI question at upload. The industry standard behind it is a DDEX field that travels with the release. Spotify's AI credits program reads it and shows a credit line. TuneCore names AI mastering as one of the things you can disclose. Answering yes does not block a release. Answering no when it was yes is the thing that gets catalogs pulled.

Which distributors signed the industry pledge

On September 14, 2026, IFPI launched the Streaming Integrity Initiative. Distributors that sign it commit to checking who their customers are, detecting fraud, and putting measures in place to identify AI-generated content and vet it for problems such as streaming fraud. 28 organizations signed at launch, among them CD Baby, Ditto Music, Symphonic, The Orchard, AWAL, RouteNote and the three major label groups (IFPI's announcement). DistroKid, TuneCore, LANDR, Amuse, ONErpm, UnitedMasters, Stem and TikTok's SoundOn are not on that list. The list was current when we checked it on September 29, 2026. Whichever distributor you use, its own AI question at upload is the one you answer.

The second label lawsuit against Suno

On September 18, 2026, Universal Music Group and Sony Music sued Suno again, this time over its v6 model. The complaint names 60,202 sound recordings and puts statutory damages at up to $9 billion. The labels argue that v6 inherits what Suno's earlier models learned. Suno said at launch that v6 was trained from scratch, and has since pointed to its users' interactions and its licensing partnerships (Variety, Bedroom Producers Blog). Nothing has been decided. The suit does not change what you disclose at upload. If you made the track with Suno, say so, and keep a note of which version you used.

What happens to undisclosed AI

Spotify removed tens of millions of spam uploads in 2026 and runs a filter for them. The filter targets volume abuse and fake artists, not honest releases. A single disclosed AI song by a real person with real artwork is not what it is looking for.

The trap

There is a growing cottage industry of tools that promise to remove watermarks or beat AI detection. Do not use them. Generator terms of service ban it, it is grounds for a distributor to drop you, and provenance systems like C2PA are reaching audio, so the mark you strip today is the one that would have proven your file's origin tomorrow. It is also the fastest way to lose royalties that were being paid.

Where mastering fits

Mastering makes a track sound finished. It does not change what the track is. Burnish keeps the provenance metadata in the files it processes, never markets itself as a way to pass detection, and prints on every report what it did to the audio and nothing else. "Release-ready" on our report means the loudness, true peak and spectrum meet streaming spec. That is a claim we can measure.