Round Hill Music Sues Suno and Anthropic for Up to $1 Billion Each Over AI Training

The AI copyright wars have a new front — and this time it’s music publishers going after not just a specialist AI music startup, but one of the world’s most valuable AI labs. On Monday, August 17, 2026, Round Hill Music filed two copyright infringement lawsuits in the US District Court for the Northern District of California: one against Suno, the AI music generation company, and one against Anthropic, the maker of the Claude chatbot. Each suit seeks statutory damages of up to $150,000 per work willfully infringed — damages that Round Hill says could “approach or conceivably exceed $1 billion” in each case.

What Round Hill Is Alleging

Round Hill Music LP, together with five affiliated funds and holding companies, owns or controls 14,364 musical compositions and 16,873 sound recordings, according to the complaints. Founded in 2010 by Josh Gruss, the private equity-backed platform manages a portfolio of music rights worth about $1.1 billion. Among the works at issue are some of the most recognizable songs in popular music: “Iris” by the Goo Goo Dolls, “Total Eclipse of the Heart” (written by Jim Steinman, performed by Bonnie Tyler), and James Brown’s “I Got You (I Feel Good)” — three songs Round Hill says have been streamed more than six billion times combined.

Each complaint attaches an exhibit of 500 compositions, described in the filings as “a prioritized representative bellwether.” But that’s just the opening hand. Round Hill says it plans to amend both cases to cover “potentially ten thousand or more” works — a scale that would push statutory damages into the hundreds of millions, and potentially past $1 billion per defendant.

The claims themselves go beyond straightforward infringement. Round Hill alleges:

• Direct copyright infringement under the Copyright Act, for reproducing its works on training servers without authorization, license, or compensation.
• DMCA violations for circumventing technological access controls — using data scraping software to bypass security measures on licensed platforms and extract protected audio and lyrics.
• CMI removal for stripping copyright management information from works, and in Suno’s case, allegedly training the models themselves to disregard CMI in both training and output “to conceal the infringement.”

The Suno complaint also names a third defendant: Bright Data, the Israeli data-scraping firm. Round Hill alleges that Bright Data supplied the proxy networks and scraping tools that let Suno extract music and lyrics from licensed platforms like YouTube Music, Deezer, and Genius, making it liable for contributory infringement. That allegation rests on reporting by 404 Media on July 15, 2026, which found that hacked Suno source code named those platforms and showed the use of Bright Data proxies on YouTube.

The Evidence Timeline

A striking feature of these complaints is how recent the key evidence is. Round Hill argues that its claims only became actionable as new facts emerged over the past few months:

• July 15, 2026 — 404 Media reports that hacked Suno source code identified YouTube Music, Deezer, and Genius as scraped platforms, and showed Bright Data proxy usage.
• June 2026 — The Atlantic publishes searchable AI training datasets that include works from Round Hill’s catalog.
• Suno’s own court filing in the major labels’ Boston case, where the company admitted its training data “includes essentially all music files of reasonable quality that are accessible on the open internet.”

The Anthropic complaint adds an unusually vivid piece of evidence. Round Hill reproduces what it says was Claude’s own response when asked to rewrite three of the songs at issue “for a modern audience.” Claude reportedly said of its outputs: “These renders keep landing very close to the originals: same structure, same hooks reused with light rewording, and full sets of verses/choruses/bridges. That’s edging past ‘inspired by’ into reproducing the copyrighted song, even when the words are swapped out.” An AI model effectively conceding that it was reproducing copyrighted songs is the kind of exhibit that juries remember.

“We Intend to Take These Cases to Trial”

Perhaps the most consequential detail is Round Hill’s posture: it says it has no intention of settling. “We are privileged to act as stewards of some of the greatest musical works of our time, and we answer to the writers and artists who created them,” said Josh Gruss. “We are not against artificial intelligence. We are against the idea that you can build a business worth billions on top of other people’s creative work and pay the creators nothing.”

Gruss emphasized that “licensing isn’t an obstacle to innovation — it is there to protect the legal owners of the raw material, their property,” and added: “We intend to take these cases to trial and to hold these companies accountable, and we will not accept a resolution that leaves songwriters and artists deprived of their rightful share of compensation.”

That’s not just rhetoric. Round Hill chose its counsel deliberately, hiring Richard S. Busch — the trial lawyer who won the “Blurred Lines” case for Marvin Gaye’s family — rather than a settlement negotiator. Busch framed the stakes bluntly: “Copyrighted works have been used to build multi-billion dollar businesses, and in the case of Anthropic a company now valued at more than a trillion dollars, while the songwriters and publishers who created the foundations of that value have received nothing. There is nothing fair about that use, and we look forward to making our arguments to the Court and in front of a jury.”

Why Anthropic, and Why Now

Anthropic might seem like an odd target for a music case — it doesn’t make a music product. But the company is already defending three music copyright suits in the same district, and the picture they paint is cumulative. Concord Music Group, Universal Music Publishing Group, and ABKCO first sued in October 2023 over roughly 500 songs. The same publishers filed a second suit in January 2026 covering more than 20,000 songs and seeking over $3 billion — a case in which CEO Dario Amodei and co-founder Benjamin Mann are named as individual defendants (Amodei moved this month to have the direct infringement claim against him dropped). BMG brought a third suit in March 2026 over 493 compositions.

The broader context matters too. Anthropic reached a $1.5 billion settlement with book authors in September 2025, demonstrating both that it will pay substantial sums to resolve training-data claims and that litigation pressure works. The company’s most recent primary funding round in May priced it at $965 billion, with subsequent estimates above $1.2 trillion — a valuation that makes a potential ten-figure damages number look less theoretical.

For Suno, the timing is similarly pointed. Warner Music Group settled and licensed Suno in November 2025; Universal and Sony remain plaintiffs in the Boston case. The Munich Regional Court ruled against Suno on July 31, 2026 in GEMA’s case, ordering it to disclose revenue and pay damages. BMG licensed Suno on August 12, 2026 — the first major rightsholder deal since Warner — even as its own Anthropic suit continued. Round Hill’s new filings land in this unsettled landscape, betting that a jury, not a licensing desk, will set the price of training on copyrighted music.

The Stakes Beyond Music

The Round Hill suits crystallize the central unresolved question of the generative AI era: does training on copyrighted works constitute fair use? Every major AI developer says yes (or at least, that the question is unresolved in their favor); most rightsholders say no. US courts have begun sketching an answer in narrow contexts, but no case has produced a definitive ruling on music training at scale. A Round Hill trial — with bellwether songs, statutory damages math, and an AI’s self-incriminating output as evidence — could be the one that does.

The cases also test the DMCA’s reach into AI training pipelines. If scraping licensed platforms through proxy networks constitutes circumvention, and if supplying those tools creates contributory liability, then the commercial scraping ecosystem that feeds many AI training runs inherits legal risk. And Anthropic’s trillion-dollar valuation ensures that whatever precedent emerges will be priced into every future negotiation between AI labs and content owners.

Round Hill’s refusal to settle makes this a pure test of legal merits rather than business convenience. Whatever a Northern California jury decides, the answer will echo far beyond the 500 bellwether songs in the complaints — it will shape how the next decade of AI training data gets assembled, priced, and licensed.

Sources are listed in the article frontmatter.