When AI Music Goes to Court, Documentation Is What's Left Standing

On September 18, 2026, Universal Music Group and Sony Music Entertainment filed their second lawsuit against the AI music company Suno, in federal court in Boston. The labels allege that Suno's newest model, v6, was built on the output of earlier models trained on their recordings without a licence — a process the complaint calls "model laundering."

Suno has said publicly that v6 was "trained from the ground up" on a new dataset, and has separately signed licensing deals with Warner Music Group, BMG, and Believe. Universal and Sony did not sign, and their filing argues that licensed inputs do not clean up a foundation they allege was tainted.

None of this is decided. These are allegations in an active case, and a court — not a blog — will determine what is true. Fact discovery in the related first case is set to close at the end of September.

But you don't need to know who wins to see what the fight is actually about. Strip away the headline damages and the legal theory, and both sides are arguing over the same thing: evidence. Who can prove what a recording is, where it came from, and when it existed.

That question used to belong to labels with legal teams. It doesn't anymore.

The evidence problem has quietly become everyone's problem

To even bring their case, the labels reportedly had to prove their recordings were inside Suno's training data. According to the filing, they did it with audio fingerprinting — matching the technical signature of their recordings against the data during discovery.

Read that again. The largest music companies on earth, with unlimited lawyers, still came down to a technical fingerprint as the thing that could establish what belonged to whom.

Now scale that down. An independent artist releasing a track in 2026 is dropping it into an environment where, by Deezer's own July 2026 estimate cited in the complaint, roughly 90,000 AI-generated tracks arrive on the platform every single day. Your original work is no longer competing only with other human work. It is competing with a flood — for discovery, for placement, for a share of the same royalty pools.

In that environment, a plausible question is coming for every creator and every catalogue: Was this made with AI? Can you show how it was made? Can you show it existed before someone else's version did?

Most independent creators have no answer. Their proof lives in an email, a project file, or nowhere at all.

What documentation actually does — and what it doesn't

This is the gap Audiverify was built for. It's worth being precise about what that means, because the honest version is stronger than the hyped one.

Audiverify does not prove you own a song. It does not register your copyright, adjudicate a dispute, or make your track "compliant" with any law. Ownership and legality are decided by courts and copyright offices, not by us. Anyone who tells you a certificate does that is overselling it.

What Audiverify does is narrower, and useful precisely because it is narrow. When you document a recording, you create:

• A SHA-256 fingerprint — a tamper-evident technical signature of the exact file. Change one note and the fingerprint no longer matches.
• An independent RFC 3161 timestamp from a third-party authority — cryptographic proof that that fingerprint existed at that moment in time. It can be verified by anyone, on their own machine, using standard tools, without taking our word for it.
• A record of what you declared about the recording — including any AI involvement, human contribution, and sound sources — captured as stated by you, not verified as true.
• A verification page and evidence pack you can share with a distributor, a platform, or a lawyer.

The distinction in that third point is the whole thing. We record what was submitted, not what is true. That sounds like a limitation. It is actually what makes the record survive scrutiny — because it never claims more than it can back up. A verified identity made these statements, at this timestamped moment, and here is the cryptographic proof of when. That is a fact. Whether the statements themselves are true is a separate question, and we don't pretend to answer it.

Why this matters more the bigger AI gets

The Suno case is one battle in a much wider shift. Regulators are moving the same direction: the EU's AI Act began enforcing new transparency provisions in August 2026, including expectations that certain AI-generated or manipulated content be marked and disclosed. Streaming platforms are rolling out AI labelling. Industry bodies have pushed for AI-made tracks to be tagged worldwide.

The common thread across the lawsuits, the regulation, and the platform policies is not "AI is banned." It is "show your work." The whole industry is being pushed toward provenance, disclosure, and reviewable records.

If you make music — with AI, without it, or somewhere in between — the useful position is not to argue about whether that shift is fair. It is to have a record. To be able to say: here is my recording, here is what I declared about how it was made, and here is independent proof of when that record was created.

You may never need it. But the entire Suno case exists because, at scale, someone eventually does.

The honest takeaway

Documentation won't win you a lawsuit, prove ownership, or make you compliant. Don't trust anyone — including us — who says otherwise.

What it does is simpler and, in a world of tens of thousands of AI tracks a day, quietly essential: it turns "trust me, this is mine and I made it in March" into a timestamped, tamper-evident, independently verifiable record that exists outside your own hard drive.

The labels needed a fingerprint to make their case. You can create one for your own work before you ever need it.

Sources & notes

Reporting on UMG/Sony v. Suno drawn from Music Business Worldwide and The Verge (Sept 18–25, 2026); Deezer platform figure and the "model laundering" language are as stated in the complaint. All statements about the litigation are allegations in an active case, not findings of fact. Nothing here is legal advice. EU AI Act transparency provisions entered enforcement from August 2, 2026. Audiverify documents what was submitted and when; it does not establish ownership or legal rights.