DistroShield / Blog
Analyze my song
Home · · Español

Reservoir sues Suno over 1,050 songs. The publisher front opens.

On October 21, Reservoir filed a copyright infringement suit against Suno in the Northern District of California, alleging the AI music company copied 1,050 Reservoir-owned compositions into its training set without permission. Reservoir filed a parallel suit against Anthropic the same day.

This is the first time a publisher outside the majors moves directly against Suno's training set. Reservoir is a mid-sized rights holder, not UMG or Sony. The songs listed in the exhibit are not catalog obscurities — they include Joni Mitchell's Both Sides Now, Johnny Cash's Ring of Fire, Madonna's Papa Don't Preach, and Evanescence's Bring Me to Life.

Reservoir alleges Suno "built its entire business model based on copyright infringement at a scale never seen before" — characterizing the operating pattern as "infringe now, settle on the cheap later." — Complaint, Reservoir v. Suno, N.D. Cal., Oct 21, 2026

A fourth front, not a repeat

The litigation pressure on Suno has arrived from three distinct directions over the last fifteen months:

Reservoir opens a fourth: publishers outside the majors, attacking the composition layer. The 1,050 songs in the exhibit are the compositions Reservoir says it owns or exclusively controls in the US. The claim is not about recordings generated by Suno — it is about whether Reservoir's compositions sat inside the training data. If the court agrees, every Suno output arguably carries a derivative-work problem traceable to those compositions, regardless of what licensing paperwork sits above the model.

Why distributors should read this closely

The second UMG/Sony suit already attacked the "partner-licensed" shield by arguing v6 inherits infringement from earlier unlicensed models. Reservoir goes further: it says the compositions themselves are in there, and names them. A publisher does not need to prove the model is bad — it needs to prove its songs were used. The exhibit is the proof.

For a distributor, the cascade is familiar. If retroactive takedowns arrive — whether driven by majors, publishers, or DSPs — they land on the pipeline that delivered the track. The upstream licensing contract between Suno and its partner is not a document the distributor holds. The signed audit of what entered the catalog is.

The acoustic signature does not care about the legal theory

Each lawsuit — majors, publishers, DSPs, distributor-side enforcement — attacks a different legal premise. The acoustic fingerprint under all of them is the same. A Suno-generated track sounds like a Suno-generated track whether a court eventually rules on training-set theory, derivative-work theory, or neither.

Our detection model listens to the audio and asks: does this sound like it came out of Suno? It does not need to know which lawsuit is live this week, and the answer does not change based on which legal theory wins. For a distributor assembling a defense of its catalog, that stability is the point.

What to do this week

  1. Pull the subset of your catalog declared as AI-assisted or fully AI since the start of 2026. Flag anything from a Suno or Suno-linked partner path.
  2. Run acoustic detection on each one. Keep the verdict, the model version that produced it, and the signed certificate.
  3. Attach the certificate to the release record. If a takedown wave arrives — publisher-driven, major-driven, or DSP-driven — you respond per-track with cryptographic evidence of what the audio was on the day of ingestion, not with a blanket removal.

The per-track signed record is the piece that converts a defensive position into an operational one. Everything above it in the stack — licensing contracts, partner attestations, AIGC declarations — is paperwork. The signed audio verdict is proof.

Every DistroShield analysis produces a signed certificate with the verdict, the acoustic score, the model version, and a SHA-256 hash. Attach it to the release record; use it in the takedown response.

Analyze a track →

Sources