Reservoir v. Suno: Inside the $186 Million AI Music Copyright Lawsuit Over Joni Mitchell, Johnny Cash and Madonna

Independent music publisher Reservoir Media has sued the AI music generator Suno, claiming the company copied 1,050 of Reservoir's songs, including Joni Mitchell's "Both Sides Now," Johnny Cash's "Ring of Fire" and Madonna's "Papa Don't Preach," to train a model that now writes complete songs for $10 a month. The case is Reservoir Media Management, Inc. v. Suno, Inc., No. 3:26-cv-12305, filed October 21, 2026, in the Northern District of California and initially assigned to Magistrate Judge Nathanael M. Cousins. Reservoir filed a separate suit against Anthropic in the same court on the same day. Billboard and Music Business Worldwide broke the story, and Billboard quoted the complaint's sharpest line: Suno was "always meant to replace the artists."

The plaintiffs are Reservoir Media Management, Inc., its UK subsidiary, and Shapiro, Bernstein & Co. Limited, which the UK entity wholly owns. They are represented by Richard S. Busch and colleagues at Adams & Reese and by Skaggs Faucette in San Francisco. A Suno representative declined to comment to Billboard.

This is not the first lawsuit against Suno. It is, however, one of the most detailed, because it is built on the company's own leaked source code, a German court loss and Suno's own licensing deals.

What Reservoir says Suno did

The complaint describes the alleged infringement in three stages.

1. Scraping. Reservoir points to Suno's answer in the major labels' case, UMG Recordings v. Suno (D. Mass.), where Suno said its training data "includes essentially all music files of reasonable quality that are accessible on the open Internet." Reservoir argues that a dataset of that scope "necessarily includes" its catalog. The complaint then relies on a July 2026 404 Media report on leaked Suno source code. Reservoir alleges the code shows Suno used proxy services from Bright Data to scrape YouTube Music, Deezer and Genius, including a YouTube Music workflow that saved 2,057,486 artist records and a Genius workflow with 6,980,653 song-page records. It also alleges Suno used the stream-ripping tools YT-DL and YT-DLP to get around YouTube's "rolling cipher" encryption.

2. Training. According to the complaint, specific Reservoir titles appear in the leaked code, including "Both Sides Now," "Ring of Fire," Ben Harper's "Diamonds on the Inside" and covers of Evanescence's "Bring Me to Life" and Joni Mitchell's "River." Reservoir also alleges the code used "stem splitting" to separate vocals from instruments, sought out a cappella versions, and paired cover recordings with originals so the model could learn to turn one version of a song into another.

3. Output. Reservoir alleges Suno's guardrails are easy to get around. It says the platform rejects "Reba McEntire" in a prompt but accepts "in the style of Reba McIntyre," and that third-party tools sell artist-specific prompt packs. It says a prompt for "a Miles Davis like Blues track with horns and a quartet" produced two tracks titled "Blue Note Quartet." The complaint also borrows output examples from other cases: the labels' allegation that Suno produced music resembling "Johnny B. Goode" 27 times, and GEMA's showing in Germany that Suno could reproduce "Daddy Cool" and "Mambo No. 5" almost note for note.

The licensing deals become evidence

The complaint's most pointed argument uses Suno's peace deals against it. Suno settled with Warner Music Group in November 2025, then announced a "global strategic alliance" with BMG in August 2026 and a partnership with Believe in September 2026. Reservoir says those deals show "there is a functioning market for licensing music and sound recordings for training data."

That matters because the fourth fair-use factor asks about harm to the market for the original, including licensing markets. Reservoir describes Suno's approach this way: "infringe now, make billions by being the first mover, and try to settle on the cheap later." It adds, "Seeking forgiveness is clearly more profitable than permission." The complaint also lists companies it says do license training data, including ElevenLabs, Musical AI and GEMA's licensing program.

The complaint cites Suno's own numbers as well: a June 2026 Series D that raised over $400 million at a $5.4 billion valuation, more than 2 million paid subscribers, about $300 million in annual revenue as of February 2026, and roughly 7 million songs generated per day as of November 2025.

The three claims and the $186 million figure

1. Direct copyright infringement (17 U.S.C. §§ 106, 501), covering both the training copies and the outputs. Reservoir seeks actual damages and Suno's profits or, alternatively, statutory damages of up to $150,000 per work willfully infringed. For 1,050 works, that's $157.5 million.
2. Circumvention of access controls under the DMCA, 17 U.S.C. § 1201, based on the alleged bypassing of logins, paywalls, CAPTCHAs and YouTube's encryption. Pleaded at $2,500 per work, or $2.625 million.
3. Removal of copyright management information under § 1202. Reservoir alleges Suno's data pipeline deliberately stripped titles, songwriter names and publisher data, which would hide the source of any copied material. Pleaded at $25,000 per work, or $26.25 million.

Together that's just over $186 million. Reservoir also asks for an injunction, an accounting of all training data, and the impoundment or destruction of copies held "in training datasets, internal repositories, model weights, and server infrastructure." Destroying model weights is the remedy AI companies fear most, and U.S. courts have rarely, if ever, ordered it.

How strong is the case?

Reservoir's advantages:

• Momentum. The complaint says Suno "has lost every step of the way," citing denied motions to dismiss elsewhere and the July 31, 2026, ruling by the Munich Regional Court in GEMA v. Suno. As the complaint describes it, that court found infringement, held that Suno needed a license under U.S. law where the training took place, and issued an injunction. The complaint also notes suits by Denmark's Koda and Canada's SOCAN.
• Recent fair-use rulings. Reservoir relies on Bartz v. Anthropic (N.D. Cal. 2025), where Judge Alsup held that keeping pirated books in a permanent central library was not fair use, even though he called training itself transformative. It also cites what it describes as a September 29, 2026, Third Circuit decision in Thomson Reuters v. Ross Intelligence rejecting fair use where an AI tool was trained to compete with the source it copied. An AI music generator whose outputs compete with human-made songs fits the "same purpose" argument better than a general-purpose chatbot does.
• The DMCA claims don't depend on fair use. If Reservoir proves that Suno got around access controls, § 1201 liability doesn't turn on whether the training was transformative. The CMI claim cites Judge Eumi Lee's 2025 ruling in Concord v. Anthropic, which let similar allegations go forward at the pleading stage.

Suno's likely defenses:

• Fair use for training. Suno will argue that learning statistical patterns from music is transformative and that its outputs are new songs, not copies. Judge Alsup's view that training is "exceedingly transformative" helps that argument, and Judge Chhabria in Kadrey v. Meta ruled for Meta on the record in front of him.
• Proof that specific works were copied. Much of the complaint is pleaded "on information and belief" and depends on leaked code that Suno hasn't authenticated. Reservoir names a handful of titles found in the code. It will have to tie all 1,050 works to Suno's datasets in discovery.
• The DMCA theory is aggressive. Courts disagree on whether robots.txt files, rate limits and terms of service are "technological measures" that "effectively control access" under § 1201. YouTube's cipher is a stronger example than a robots.txt file.
• Venue and consolidation. Suno is based in Cambridge, Massachusetts, and the labels' main case is pending there. Reservoir argues for San Francisco based on Suno's SoMa office and Oracle servers in San Jose. Expect a motion to transfer, or an effort to coordinate this case with Round Hill v. Suno and Gerencia 360 v. Suno, both already before Judge Rita F. Lin in the same district.

The most likely result is the one Reservoir all but invites: a license. Warner, BMG and Believe already have deals. Each new publisher suit raises the cost of staying unlicensed, and Suno's $5.4 billion valuation gives it the money to pay.

The companion case against Anthropic

Reservoir's second complaint, Reservoir Media Management, Inc. v. Anthropic PBC, No. 5:26-cv-12308 (N.D. Cal.), makes similar claims over song lyrics used to train Claude. An Anthropic spokesperson told Billboard, "We disagree with the allegations and intend to defend this case vigorously." It joins the long-running Concord Music Group v. Anthropic lyrics case before Judge Lee, where fair-use arguments are now being heard. Follow it on IP Feed at /case/47195.

What this means for filmmakers

Generative music is reaching indie film quickly. A Suno track costs less than a library cue and takes seconds to make. This lawsuit shows why that shortcut is risky for anything you plan to distribute.

• Errors and omissions insurers will ask about AI. E&O carriers and distributors want a clean chain of title for every cue. A track from a model accused in court of training on unlicensed music, with $186 million at stake in this case alone, is hard to clear. Keep records of which tool generated each cue and under which terms.
• AI-generated music may not be copyrightable. The U.S. Copyright Office's position is that purely AI-generated material isn't protected by copyright. A Suno score may not be something you own or can license to a buyer, even if the tool's terms say otherwise.
• Output can still infringe. The complaint's evidence about outputs, such as "Johnny B. Goode" lookalikes and artist-style prompts, means an AI cue could resemble a real song closely enough to draw a claim. That's a problem in sync licensing, where a single song can tie up an entire film.
• Licensed AI is coming, so watch the terms. Suno's newest model was trained under the Warner, BMG and Believe deals. Licensed models are safer, but read whether the license covers sync in film and TV, worldwide, in perpetuity. Those are the rights a sales agent will ask about.
• For composers, this is leverage. If courts treat training licenses as a real market, the scores and songs you own become licensable assets. Register your copyrights. Statutory damages, which drive this lawsuit's $157.5 million figure, are available only for works that were timely registered.

Follow the Suno cases on IP Feed: Reservoir v. Suno, Poseidon Wave Media v. Suno, and the UMG v. Suno discovery dispute in San Francisco. Track the parties at Suno, Reservoir Media, Bright Data, Round Hill Music and Anthropic. The docket is on CourtListener, and we'll cover major rulings here.

This post is editorial commentary on public court filings and news coverage, not legal advice. Reservoir's allegations, including those based on leaked source code and press reports, have not been proven, and Suno has not yet responded in court. Quotes are from the complaint in No. 3:26-cv-12305 (N.D. Cal.) unless attributed to Billboard or Music Business Worldwide. Descriptions of other rulings, including GEMA v. Suno and Thomson Reuters v. Ross, reflect the complaint's characterization and press reports.