Music publisher Reservoir sued Suno and Anthropic on October 21 in federal court in California, and the Reservoir Suno lawsuit now sits inside a growing fight over whether AI companies can train on copyrighted songs without permission. The filing came days after a federal judge heard Anthropic's fair-use argument in a separate lyrics case, and that judge has not ruled.

The complaint against Suno, as described by Music Business Worldwide, alleges that the AI music company copied 1,050 of Reservoir's compositions to train its models without permission. That outlet says it obtained the complaint, and it places the filing in the U.S. District Court for the Northern District of California on Wednesday, October 21. No other outlet had confirmed those details in our review, so treat them as provisional until court records or additional reporting surface.

What the Reservoir Suno lawsuit alleges, and what is still unknown

The Music Business Worldwide report describes a copying-without-permission theory. Its headline names songs by Joni Mitchell, Johnny Cash and Madonna among the works at issue, which is the most concrete public detail so far.

Reservoir also filed a separate suit against Anthropic in the same court on the same day. Public reporting confirms that filing but does not describe its claims, the number of works involved or the damages sought. Neither Suno nor Anthropic had publicly responded in the sources we reviewed, so the defense side of both cases is still unwritten.

Those gaps matter. A works list would show how much of Reservoir's catalog is actually in dispute, and a damages figure would show how much the publisher expects the claim to be worth. Until those appear, the suits are best read as a signal of strategy rather than a measure of exposure.

Why the Concord lyrics case sets the terms

The Concord case is brought by UMG, Concord and ABKCO, who argue that Anthropic's Claude reproduces their song lyrics, sometimes verbatim. On October 21, Judge Eumi Lee heard Anthropic's motion for summary judgment on fair use in San Jose. Summary judgment asks a judge to rule on the legal arguments without a trial, when the key facts are largely settled.

At the hearing, the publishers cited almost 120,000 prompts asking Claude for “new” song lyrics, according to AI Music Preneur. Anthropic's position is that training on the lyrics of 499 songs is fair use. The same litigation tracker says Concord is the first AI-training fair-use summary judgment argued in a federal district court since the 2025 rulings in Bartz v. Anthropic and Kadrey v. Meta. That makes Judge Lee's eventual decision a near-term reference point for everyone else in the docket.

Judge Lee has not ruled, and no date has been set, according to FourWeekMBA.

Anthropic's earlier settlement offers the clearest dollar benchmark in this docket. In September 2025, the company agreed to pay $1.5 billion to settle a class action by book authors. That deal did not answer the fair-use question, and it involved books rather than songs, but it shows the price tag that can follow a copyright fight over training data.

Where the Ross ruling fits into the fair-use fight

The most consequential legal development in these filings may come from a case that is not about music. In Thomson Reuters v. Ross Intelligence, the Third Circuit rejected a fair-use defense for AI training that competes with the source product. The litigation tracker lists the decision as affirmed on September 29, 2026. Ross Intelligence built a legal research tool, and the AI Copyright Lawsuits tracker describes the ruling as the first US decision to reject fair use for AI training that directly competes with the source product.

The ruling is already showing up in court. On October 23, news outlets suing OpenAI and Microsoft cited it in their opposition to the defendants' summary-judgment motions, according to MLex. The decision is also being cited in the newest filings.

The hard part is that other courts have not adopted the same rule, so the decision's reach remains disputed. The news publishers argue it helps them, while defendants may argue it is distinguishable. For music, the core question is whether a lyric or song generator competes with the songs it learned from. That test is harder to apply than it sounds, and a ruling in the Concord case could shape how the question gets argued even though it would not bind courts elsewhere.

How the music publisher suits stack up

Reservoir is not the first music publisher to sue over AI training this year. The sequence so far looks like this:

  • March 2026: BMG sues AI music firms over 493 compositions.
  • August 28, 2026: Sony Music Publishing and Warner Chappell Music file their own suit.
  • October 21, 2026: Reservoir sues Suno and, in a separate filing, Anthropic, both in the same federal court.

The pattern suggests publishers are filing one case after another rather than waiting on a single verdict. That approach spreads the legal load across several sets of facts, and it forces AI companies to defend more than one theory at a time.

One tracker lists 144 AI copyright cases, and many are still active, according to the AI Copyright Lawsuits (2026) tracker. The music cases are one slice of that total. On October 21, Postmedia, Torstar, The Globe and Mail and The Canadian Press sued Anthropic in Canada. That is a non-US suit and is not part of the Reservoir case, but it means Anthropic faces publisher claims on both sides of the border.

News groups are pressing their own claims too. According to Press Gazette, USA Today Co. has sought $250 million from OpenAI over articles across 19 publications. Trackers disagree on the filing date for that complaint, with one listing October 8 and another October 9.

Not every AI case involves copyright. Rhode Island sued Meta over scam ads and its AI ad tools, and Texas sued OpenAI and Sam Altman over ChatGPT. Those cases show state governments pursuing AI companies on consumer and advertising grounds, a separate track from the training-data fight.

What songwriters, publishers and AI users can do now

No official guidance accompanies these steps. The sources reviewed for this story did not include any, so treat the following as general practice rather than legal advice.

  • Songwriters and publishers should confirm which compositions they own or control, and keep registration records where they exist. A clear catalog list makes it faster to check whether your works appear in a complaint.
  • Talk to a copyright attorney before deciding whether to file, license or join a claim. A lawyer can review your contracts and ownership splits.
  • Businesses using AI music tools for commercial work should read the vendor's terms and indemnity language. Check whether the vendor promises to defend claims tied to its outputs, and what that promise covers.
  • Ask vendors whether they have licensed their training data. The answer affects how exposed your projects may be if courts side with rights holders.
  • Keep records of which tools produced which assets, so you can respond quickly if a claim reaches your work.

What to watch next

The biggest open item is Judge Lee's decision in Concord v. Anthropic, which has no date. The tracker also lists a hearing in In re Mosaic LLM Litigation on October 30, 2026, and oral argument in UMG/Concord/ABKCO v. Anthropic II on November 4, 2026.

Watch whether Suno and Anthropic respond to the Reservoir complaints, and whether the claims against Anthropic become public. Watch too whether defendants in the news cases persuade judges that the Ross decision is distinguishable. Each answer will shape how much leverage publishers and news organizations carry into talks with AI developers.

A fair-use ruling in either direction would strengthen one side of the next round. A ruling for Anthropic would make training on lyrics look more defensible in court, while a ruling against it would likely push more rights holders toward licensing talks and new filings.