Suno admits v6 was trained on user creations. What that means for distributors who accepted licensed AI.

Universal and Sony sued Suno a second time this month, targeting the v6 model specifically. Their argument: even if v6 was trained on catalog licensed from Warner and other partners, it was also trained on the outputs of Suno's previous models — models the majors say were built on their unlicensed copyrights. A copy of a copy of a copy still carries the original.

Suno's own spokesperson response effectively confirmed the premise:

"Creations from our community" are, by definition, tracks generated with the earlier Suno models. Those earlier models are exactly what UMG and Sony allege were built on scraped catalogs. The chain is now on the record from Suno itself.

Why this matters for distributors that opened doors to Suno

Two weeks ago we covered Believe reversing its Suno block and opening TuneCore's front door to partner-licensed AI. Other distributors have signaled similar pivots. The premise across all of them is the same: "partner-licensed" is a legal safety line. If the model was trained on licensed material, tracks generated by it are safe to distribute.

The second UMG/Sony lawsuit — and Suno's own reply — attacks that premise directly. If a court accepts that v6 inherits infringement from earlier unlicensed models, then every "partner-licensed" v6 track already sitting in a distributor's catalog carries retroactive exposure. Legal cleanliness at the model layer no longer implies legal cleanliness at the output layer.

The distributor is not the one who signed with Suno. But the distributor is the one whose pipeline delivered the track to Spotify, Apple, Deezer. If retroactive takedowns arrive, they land on the distributor's side of the ledger.

Legal framing changes. Acoustic signature does not.

A track generated with Suno v6 sounds like a Suno v6 track — regardless of whether the model was fully licensed, partially licensed, or trained on scraped catalogs. The acoustic fingerprint of the generator persists across every legal reframing above it.

Our detection model is trained to recognize that fingerprint. It doesn't ask what license the model has this week. It listens to the audio and asks: does this sound like it came out of Suno? Same question, same answer, whether the model in question is v3.5, v4, v5, or v6. Same question, same answer, whether the track was made by an anonymous free user in 2024 or by a partner-licensed release in 2026.

For a distributor holding a catalog of AI tracks accepted under "partner-licensed" attestations, that means one thing: the audit that establishes what you actually delivered is possible even after the fact. You do not need to trust the artist's declaration or the licensing partner's paperwork. You can check the audio.

What to do this week if you distribute AI music

If your pipeline started accepting partner-licensed Suno tracks in the last few months, three checks are worth running now, before any court decides anything:

1. Identify which releases in your catalog were declared as AI-assisted or fully AI, and which came from Suno or a Suno-linked partner path.
2. Run acoustic detection on those releases. Store the verdict, the model version that made the call, and a signed record for each one.
3. Keep the signed evidence attached to the release record, so if a majors-driven takedown wave arrives you can respond with per-track documentation instead of a blanket removal.

The signed record is the piece a distributor can actually escalate with. It's not an opinion, it's a cryptographic proof of what the audio was on the day it entered the catalog.

Sources

• Music Business Worldwide — reporting on the second UMG/Sony lawsuit and Suno's spokesperson statement on v6 training
• 404 Media — reporting on AI-training-AI patterns across other model providers
• DistroShield blog — Believe reverses on Suno: what TuneCore artists need to know
• DistroShield blog — Judge cleared Universal and Sony to accuse Suno of pirating YouTube