Reservoir Media Management, the US independent music publisher behind catalogues that include Joni Mitchell's "Big Yellow Taxi" and Johnny Cash's "Ring of Fire", sued Suno on October 21, 2026 in the US District Court for the Northern District of California, accusing the AI music generator of copying 1,050 compositions to train its models and seeking $157.5 million in statutory damages on its main copyright count alone.

In Short

A music company that owns the rights to famous songs says Suno, an app that makes whole songs from a text prompt, copied those songs without paying so it could teach its AI how music works. The company wants up to $150,000 for each of 1,050 songs, plus smaller amounts for allegedly breaking through website protections and stripping out songwriter credits. If the claims succeed, the case would add to pressure on AI music tools to license their training data, which matters for anyone using such tools to make soundtracks, jingles or ads.

The filing

The complaint, case number 3:26-cv-12305, runs to 28 pages and was signed by Richard S. Busch of Adams & Reese, LLP in Nashville, alongside Skaggs Faucette LLP in San Francisco. Three entities are named as plaintiffs: Reservoir Media Management, Inc., a Delaware corporation headquartered at 200 Varick Street in New York; Reservoir Media Management (UK) Limited, a wholly owned subsidiary formed in England and Wales; and Shapiro, Bernstein & Co. Limited, which the UK entity in turn owns outright. The parent company, Reservoir Media, Inc., trades on Nasdaq under the ticker RSVR.

The defendant is Suno, Inc., a Delaware corporation with its principal place of business at 17 Dunster Street in Cambridge, Massachusetts. Three causes of action are pleaded: direct copyright infringement, circumvention of technological protection measures under 17 U.S.C. section 1201, and removal or alteration of copyright management information under section 1202. Reservoir demands a jury trial.

Reservoir describes itself in the filing as "the first female-founded and -led publicly traded independent music company in the United States." Its catalogue, according to the complaint, covers "tens of thousands of musical compositions and master recordings."

The works at issue are listed in Exhibit A, which the complaint says contains 1,050 compositions in which Reservoir "owns 100 percent or exclusively controls the United States copyrights." Each, according to the filing, is registered with the US Copyright Office. The exhibit is referenced throughout but sits outside the 28-page body of the complaint. The body names a selection: Mitchell's "Big Yellow Taxi" and "Both Sides Now", Billy Strayhorn's "Lush Life", Madonna's "Papa Don't Preach", Miles Davis's "So What", "In the Mood" as recorded by Glenn Miller and His Orchestra, four Evanescence songs including "Bring Me to Life", Drowning Pool's "Bodies", and Finger Eleven's "Paralyzer", among others.

Bright Data, the web data company whose proxy services feature heavily in the factual allegations, is not a defendant. The complaint describes Bright Data Ltd. and its wholly owned subsidiary Bright Data, Inc. as suppliers of tools Suno allegedly used, but the only party Reservoir asks the court to rule against is Suno.

What Reservoir says Suno did

Training on "essentially all music files"

The starting point for the training allegation is a statement Suno made in separate litigation. In its answer in UMG Recordings, Inc. v. Suno, Inc., the case brought by the major record labels in the District of Massachusetts in June 2024, Suno stated that its "training data includes essentially all music files of reasonable quality that are accessible on the open Internet, abiding by paywalls, password protections, and the like."

Reservoir argues that a dataset of that breadth "necessarily includes" its works. Its songs, the filing says, are available on Spotify, Apple Music, YouTube, Amazon Music, SoundCloud and Tidal, all licensed by Reservoir and all requiring "some form of account registration, subscription, or other access credentials to use." Given "Suno's self-described training methodology," the complaint says, "it is a near certainty that Reservoir's Works were copied into Suno's training sets."

There is a tension between the two documents worth noting. Suno's own description, as quoted, says the training data was gathered while "abiding by paywalls, password protections, and the like." Reservoir's complaint alleges the opposite: that Suno accessed paywalled platforms such as Tidal "by scraping and circumventing technological access controls." How those two accounts are reconciled will matter for the section 1201 claim.

Leaked source code

The most specific factual material in the complaint comes from what it calls "leaked source code from Suno," as reported by 404 Media. According to the filing, the code shows Suno used proxy services acquired from Bright Data and scraped "millions of songs and lyrics" from YouTube Music, Deezer and Genius.

The numbers given are precise. A YouTube Music workflow in the code, according to the complaint, shows 2,057,486 artist records saved as output. For each artist, the code requests the artist's song playlist "without a track limit," falling back to song results already available when no playlist identifier exists. Reservoir characterises this as "a systematic effort to collect across artists' catalogs." A separate Genius workflow records 6,980,653 successful song-page processing records, which the filing says confirms "that not less than this number of extracted lyrics exist in Suno's code."

A file named "youtube_music", the complaint adds, holds over two million music clips amounting to approximately 113,879 hours of audio - which the filing equates to "thirteen years" of music.

Reservoir then names specific works it says appear in the code. "Both Sides Now" by Joni Mitchell, "Diamonds on the Inside" by Ben Harper, "Ring of Fire" as performed by Johnny Cash and "Jukin' City" are listed as appearing in the leaked source code. Another file allegedly carries song metadata for "8th of November" by Big & Rich, Mitchell's "Good Friends", a cover of Mitchell's "River", "Eyes on You" and "Over" by Kings of Leon, and a cover of Evanescence's "Bring Me to Life." For the remaining works, the complaint relies on information and belief: "all of the Works exist in some form or fashion in Suno's code, training sets, and have been unlawfully reproduced by Suno."

Stream ripping and the rolling cipher

On YouTube specifically, the complaint describes a technical protection it calls the "rolling cipher." YouTube, according to the filing, maintains two URLs for every video: a visible page URL and a hidden file URL for the media itself. The file URL "is encrypted using a complex and periodically changing algorithm" designed to impede external access to the underlying files. The video player decrypts or descrambles it to locate and play the content.

Reservoir alleges that Suno used two tools, YT-DL and YT-DLP, to circumvent that encryption, a practice the complaint calls "stream ripping." The filing describes this as "the key means by which Suno engages in its first infringing act of reproducing the Works." Public knowledge of the scraping, it says, dates to on or about September 2, 2025, when the ICMP shared private datasets with Billboard.

The same rolling-cipher finding carried weight in Germany. When the Munich Regional Court ruled against Suno on July 31, 2026, it found that Suno obtained recordings by circumventing YouTube's rolling cipher, which undermined Suno's reliance on the text and data mining exception.

Bright Data's tools

The complaint spends two paragraphs on Bright Data's product line. It describes "residential and datacenter proxy networks that rotate IP addresses to evade rate limits and IP bans," a "Web Unlocker" service "designed to defeat CAPTCHAs, JavaScript challenges, and bot-detection systems," and a "Scraping Browser" that renders pages "without the headers and footers to circumvent client-side protections."

That last detail is the hinge of the copyright management information claim. "CMI regularly exists in the headers and footers of files containing musical works," the filing says. "Making the deliberate choice to exclude headers and footers is identical to making the deliberate choice to remove CMI from these files." The complaint gives no technical detail on which headers or footers it means - whether page elements of a lyrics website or metadata embedded in audio files - and the two are different things.

Reservoir alleges Bright Data's services were used to defeat "authentication gates, subscription paywalls, rate limiting, CAPTCHA challenges, and robots.txt access restrictions." The access controls listed in the section 1201 count are broader still: login credentials and session tokens, paid subscription paywalls, CAPTCHA and bot-detection challenges, IP-based rate limiting, JavaScript rendering requirements, and "Terms of Service and robots.txt directives restricting automated access."

Whether a robots.txt file or a terms-of-service page can count as a technological measure that "effectively controls access" under the statute is a contested legal question. A robots.txt file is a request to crawlers, not a lock. The theory also meets a recent obstacle in the same district: in July 2026, Chief Judge Yvonne Gonzalez Rogers dismissed Google's section 1201 claims against SerpApi, finding that Google's SearchGuard system did not effectively control access to a copyrighted work for most search results, and that Google had not pleaded the copyright owner's authorisation for the rest.

Outputs, guardrails and the V6 model

The complaint does not stop at training. It alleges that Suno's outputs themselves infringe, and it builds that claim from several sources.

The first is Suno's chief executive. The filing quotes Mikey Shulman, in a video titled "Suno Generates Beautiful Songs with Oracle Cloud", saying: "Our models are taught to just continue a piece of music. And with that, and a little bit of extra sauce for getting these things to start properly, we can go from an idea in somebody's head, to an entire three minute song very easily." Reservoir reads this as an admission that the model extends existing music rather than generating songs "from whole cloth."

The second is guardrails. According to the complaint, the platform currently refuses the prompt "Reba McEntire" but accepts "in the style of Reba McIntyre" - a misspelling the filing marks with "[sic]" - to produce a female-vocal country song in her style. Third-party services such as HookGenius, the filing says, advertise artist-specific prompts to get around Suno's filters.

Suno's V6 model, launched September 9, 2026, now accepts artist-name prompts and "redirect[s]" them "toward descriptive musical characteristics," according to the complaint, which cites an August 6, 2026 Suno blog post by Shulman. Reservoir argues this shows that "Suno was always designed to model after pre-existing works."

User uploads form a third strand. Suno has accepted audio uploads since 2024, the complaint says, and its terms grant the company a "perpetual, irrevocable" licence to train on user-uploaded content. The announcement of new audio and lyrics safeguards for V6 is described in the filing as "at least a tacit admission that the previous safeguards were not effective enough."

The complaint then describes a test. When prompted to "give me a Miles Davis like Blues track with horns and a quartet," Suno generated two lyric-less tracks, both named "Blue Note Quartet," which Reservoir calls "unlicensed derivative works of Miles Davis music." The filing does not describe a melodic comparison between those outputs and any specific Davis composition in Exhibit A.

For more concrete output evidence, Reservoir borrows from other cases. It cites the UMG complaint's allegations that Suno generated music resembling Chuck Berry's "Johnny B. Goode" 27 times, Bill Haley & His Comets' "Rock Around the Clock" in ten outputs, Jerry Lee Lewis's "Great Balls of Fire" in ten, B.B. King's "The Thrill Is Gone" in six, and "Sway" in eleven. None of these are Reservoir works. It also cites GEMA's lawyers reproducing "Daddy Cool" by Boney M. and "Mambo No. 5" "almost note for note" in the German case.

Stem splitting and cover pairing

Two technical allegations about the training pipeline stand out. The complaint says the leaked code "deliberately enabled 'stem splitting'" - separating recordings into tracks for vocals, drums and individual instruments - and that Suno "purportedly also deliberately sought out a cappella versions of the songs" to improve vocal and lyric generation.

The second concerns paired training examples. According to the filing, "The Suno code puts a cover recording followed by the 'main' (original) recording of the same song when creating a training data instance. That gives the Suno model the cover as reference audio while it learns to predict the main recording." Reservoir suggests this could be used to turn an acoustic performance into a full-band version, or a ballad into a dance track.

The money

Suno's commercial figures appear throughout the complaint, mostly drawn from press coverage. According to the filing:

  • Suno became widely available with version 3 on March 24, 2024; version 5.5 launched in March 2026 and version 6 on September 9, 2026.
  • Over 10 million people had created music files with Suno as of 2024, according to a Suno blog post from May 21, 2024.
  • Users were generating about 7 million songs per day as of November 2025.
  • As of February 26, 2026, Shulman claimed more than 100 million people had used Suno, with over 2 million paid subscribers and $300 million in annual revenue, citing Forbes.
  • Suno closed a Series D in June 2026, raising over $400 million at a $5.4 billion valuation, according to Music Business Worldwide.
  • Revenue is projected to reach $1 billion by 2028, citing Billboard.
  • Full use of the platform costs $10 per month, under a "freemium" model.

The complaint's damages arithmetic is explicit. Under the Copyright Act, Reservoir seeks actual damages and Suno's profits or, in the alternative, statutory damages of up to $150,000 per work willfully infringed. "Reservoir identifies 1,050 Works, meaning that the total statutory damages sought in this case is $157,500,000.00," the filing states.

On the section 1201 count, it seeks up to $2,500 per violation, and multiplies by the number of works: $2,625,000. On the section 1202 count, up to $25,000 per violation, again per work: $26,250,000. By PPC Land's calculation, the three figures together come to $186,375,000. That is a ceiling built on willfulness findings and the maximum per-work award; statutory damages are set by a jury or judge within a range, and the lower end for ordinary infringement is far smaller.

The prayer for relief goes beyond money. Reservoir asks for a permanent injunction, and for an order directing Suno to deliver for impoundment or destruction all unauthorised copies of its works, "including copies retained in training datasets, internal repositories, model weights, and server infrastructure." It also seeks "a complete accounting of all training data, scraping activities, and datasets involving Reservoir's copyrighted Works." The request to destroy model weights, if granted in any form, would reach the trained model itself rather than only the stored audio.

Why San Francisco

Suno is based in Massachusetts, where the UMG case is pending. The complaint devotes a full section to explaining why the Northern District of California is the right forum.

Reservoir relies on three connections. First, servers: Suno uses Oracle Cloud Infrastructure "to train its proprietary machine learning models," and Oracle's servers, according to the filing, include a location in San Jose. Second, an office: the complaint cites Axios reporting from April 3, 2026 that Suno opened a San Francisco office in the SoMa district in late March 2026, quoting co-founder and chief technology officer Georg Kucsko saying the move "will be critical as we continue to scale." The filing lists August 2026 job postings in the city, including Lead Machine Learning Data Operations and Senior Counsel. Third, investors: the complaint names Matrix Ventures for the seed and Series A rounds, Lightspeed Ventures for Series B, Menlo Ventures for Series C and Bond Capital for Series D, all with Northern District offices. That passage refers to "Sonos five fundraising rounds," an apparent typographical error for Suno.

The choice of forum places the case alongside at least two other suits against Suno in the same court. The complaint cites Round Hill Music LP v. Suno, Inc., case 3:26-cv-8507, where an amended pleading filed October 19, 2026 discusses a Suno output "lifting nearly verbatim the lyrics of 'I Got You (I Feel Good)' by James Brown." It also cites Gerencia 360 Music, Inc. v. Suno, Inc., case 3:26-cv-09248, where a filing on October 18, 2026 discusses an output based on English-translation lyrics of "Quince Besos" by Cornelio Vega y Su Dinastia. Both carry the same judge's initials, RFL.

Northern California is also where the most-cited AI training rulings have come from. Reservoir leans heavily on one of them.

The fair use argument

Suno has defended training on copyrighted music as fair use in other cases. Reservoir's complaint tries to close that door in advance with two precedents.

The first is Bartz v. Anthropic PBC, decided in the same district on June 23, 2025. The complaint describes the ruling as holding that "indefinite retention of unlicensed copies of copyrighted works" in a general-purpose archive is not fair use, and quotes Judge William Alsup: "There is no carveout, however, from the Copyright Act for AI companies." Reservoir argues that "Suno has engaged in the same conduct."

The complaint does not mention the other half of that ruling. As PPC Land reported at the time, Judge Alsup found that training Anthropic's models on copyrighted books was fair use and called it "exceedingly transformative," while rejecting the defence for a central library built from pirated books. Anthropic later agreed to a settlement of at least $1.5 billion covering roughly 500,000 works, about $3,000 per work. Reservoir's framing centres on the library and retention part of the ruling - the part authors won - and argues Suno's conduct falls on that side of the line.

The second precedent is newer. The complaint cites Thomson Reuters v. ROSS Intelligence, decided by the US Court of Appeals for the Third Circuit on September 29, 2026, which held that ROSS's use of Westlaw material to train a competing legal research tool was not fair use. The filing links that reasoning to the Munich court's treatment of US fair use in the GEMA case, quoting Music Business Worldwide's description that the German court found "outputs served the same purposes as the originals rather than a new one, and that the use was commercial."

Reservoir also argues that licensing alternatives exist. It names ElevenLabs, Musical AI, Symphonic, Soundverse, GEMA (through PLAI), GCX/Rightsify and Troveo as companies or collective bodies that "seek permission and pay rights holders to create data sets for AI training." It points to public domain material too: every US work written before 1931, and sound recordings before 1926, according to the filing.

Suno's deals as evidence

The complaint turns Suno's licensing agreements against it. According to the filing, Suno settled with Warner Music Group in November 2025 - the Warner press release cited is dated November 25, 2025 - agreeing to "deprecate[]" its then-current models and launch V6. Suno followed with a BMG alliance announced August 12, 2026 and a partnership with Believe announced September 8, 2026.

The filing quotes Shulman on the Warner arrangement, as reported by Billboard: "I don't think of what we did as a settlement. I think about this much more like a partnership . . . . It is much more long term." Reservoir's reading is blunt: "Suno has all but admitted that there is a functioning market for licensing music and sound recordings for training data."

A quote from an unnamed Suno investor, drawn from a March 2024 Rolling Stone feature, is used to argue intent: "Honestly, if we had deals with the labels when this company got started, I probably wouldn't have invested in it. I think that they needed to make this product without the constraints."

The licensing pattern mirrors what happened at Suno's main rival. Udio settled with Universal Music Group in October 2025 and agreed to launch a licensed platform in 2026, a deal that drew questions over consent and compensation from the Music Artists Coalition within days.

Outside the US, the complaint lists GEMA's July 31, 2026 win in Munich, a suit by Koda, Denmark's collecting society, and a SOCAN suit filed in the Federal Court of Canada on September 2, 2026. The filing attributes the Munich findings to Judge Elke Schwager, including that it was "undisputed that Suno obtained the training data through stream-ripping."

What remains unproven

A complaint is one side's account. Many of its central allegations are made "upon information and belief," the standard formula for claims the plaintiff expects to prove through discovery. That includes the use of YT-DL and YT-DLP, the use of Bright Data's tools to extract Reservoir's works specifically, and the assertion that every one of the 1,050 works exists in Suno's training sets.

The direct evidence tying named Reservoir works to Suno is drawn from leaked code reported by 404 Media, which the complaint does not attach. The output evidence comes largely from other plaintiffs' songs. The filing states that "discovery will undoubtedly show that Suno created additional outputs which infringe upon all the Works," which is a prediction rather than an allegation of a specific output.

Suno's response is not part of the filing, and the document contains no statement from Suno about Reservoir's claims.

Why this matters for marketers

AI-generated music has moved quickly into advertising work. Brands and agencies use text-to-music tools for social video beds, podcast spots and short-form ad creative, where licensing a recognisable track has always been expensive. Every new suit against a generation platform raises the question of what rights a commercial user actually receives when a tool's training data is in dispute. Reservoir's request for destruction of model weights, and the GEMA injunction already granted in Germany, point to a risk that a model in use today may be withdrawn or replaced, as Suno did with its pre-V6 models after the Warner deal.

Distribution is shifting too. YouTube Music in October 2026 began labelling songs made primarily with generative AI, and Deezer reported in June 2026 that about 75,000 AI-generated tracks a day, over 44% of its daily deliveries, were arriving on its service, with tools like Suno and Udio among the sources its detector identifies.

There is a second audience for this case: the data infrastructure industry. The complaint treats a residential proxy network, CAPTCHA-solving services and headless browsers as instruments of circumvention. The same categories of tools underpin rank tracking, ad verification, price monitoring and competitive intelligence across digital marketing. Reservoir has not sued Bright Data, but a ruling that accepts its theory of section 1201 - that defeating rate limits, bot detection or robots.txt directives can amount to circumvention under the DMCA - would reach well beyond music. The SerpApi ruling in July 2026 went the other way on a different set of facts, which leaves the question open in the Northern District.

Timeline

Summary

Who: Reservoir Media Management, Inc., Reservoir Media Management (UK) Limited and Shapiro, Bernstein & Co. Limited, represented by Adams & Reese and Skaggs Faucette, against Suno, Inc., the Cambridge, Massachusetts maker of an AI music generator. Bright Data is named in the allegations as a supplier of scraping tools but is not a defendant.

What: A 28-page complaint alleging direct copyright infringement, circumvention of technological protection measures under 17 U.S.C. section 1201 and removal of copyright management information under section 1202, covering 1,050 compositions. Reservoir seeks up to $157.5 million in statutory damages on the copyright count, a further $2.625 million and $26.25 million on the two DMCA counts, an injunction, destruction of copies including in model weights, and an accounting of training data.

When: The complaint was filed on October 21, 2026.

Where: The US District Court for the Northern District of California, case 3:26-cv-12305, where at least two other music publishers' suits against Suno are pending.

Why: Reservoir alleges Suno trained its models on its songs without a licence, relying on Suno's own statement that its training data includes "essentially all music files of reasonable quality," on leaked code listing specific Reservoir works, and on alleged stream ripping and proxy-based scraping. Suno has since signed licensing deals with Warner, BMG and Believe, which the complaint presents as evidence that a licensing market exists for the works it did not license.