I opened the complaint at midnight and felt the old legal drumbeat return. You can hear the music industry tightening its grip—and the language in this filing bites. I want to walk you through what the suit actually says and why it matters to you.
In a late-night filing in Northern California: The suit lands with blistering rhetoric
Sony Music Publishing and Warner Chappell dropped a lawsuit that reads like a war room memo. They named Anthropic and its co-founders Dario Amodei and Benjamin Mann, and the complaint accuses the company of “one of the largest and most blatant ongoing thefts of intellectual property in history.”
The filing alleges that Anthropic scraped authorized lyric services such as MusixMatch and LyricFind, then let Claude—the company’s chatbot—reproduce full lyrics on request.
What are music publishers accusing Anthropic of?
They claim unauthorized scraping of licensed lyric repositories and the verbatim output of protected text from Claude. That’s a narrow-sounding charge—mostly about song lyrics—but the plaintiffs are packaging it as a pattern, not an accident.
On the procedural stage outside the courtroom: How the complaint leans on earlier disclosures
Files from the Bartz v. Anthropic litigation are playing a supporting role in this new case. You might remember Bartz: Anthropic agreed to pay $1.5 billion (≈ €1.4 billion) to settle allegations that it had downloaded thousands of books from Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi).
The publishers use those revelations to sketch a pattern of conduct—torrenting e-books, in their telling, as a curious backdoor to training data. It reads like a Trojan horse argument: the book case proves method, and the lyric case proves motive.
At the core of the factual claims: Examples that read like show-and-tell
The complaint cites specific songs—“Ain’t No Mountain High Enough,” “All I Want for Christmas Is You,” “Eye of the Tiger,” “Here Comes Santa Claus,” “Paper Rings”—as instances Claude returned verbatim. That’s the same playbook the New York Times used against OpenAI, Microsoft, and Perplexity: highlight exact reproductions and argue commercial harm.
Can chatbots reproduce song lyrics verbatim?
Yes—publishers say they can and that Claude has done it. Anthropic’s critics point to logs and to the accessibility of licensed lyric databases as proof that the model’s training or retrieval processes included copyrighted text.
In the veins of industry strategy: Why this suit feels like both leverage and theater
There’s legal theater here, and there’s clear leverage. The plaintiffs seek $150,000 (≈ €140,000) per alleged infringement—numbers that can balloon into the billions if accepted at face value. That kind of math pushes defendants hard toward settlement, or a long, expensive fight.
I’ve seen cases framed this way before: start with powerful examples, tie them to a broader pattern, and force a choice—pay, litigate, or redesign the product. You feel the pressure like a needle on the record player—sometimes it skips, sometimes it screams.
How large could damages be?
The $150,000 per-infringement statutory measure is designed to sting. Multiply that across dozens, hundreds, or thousands of alleged infringements and the headline figure becomes staggering, which is precisely why publishers bring it up now.
On the likely paths forward: What I’m watching in the next moves
Expect three streams: discovery that digs into training corpora and contracts with sites like MusixMatch and LyricFind; motions over whether reproducing lyrics from a model counts as actionable harm; and settlement talks that will test how much Anthropic is willing to pay to avoid precedent.
There’s also the public-relations dimension. Sony and Warner can shape the narrative; Anthropic will argue industry norms or fair use defenses. OpenAI’s and Microsoft’s fights with publishers and the New York Times offer roadmaps—but no guarantees.
I’ll be watching the discovery filings for any smoking-gun evidence that lyrics were included in training data and for technical details about Claude’s retrieval behavior. If you follow legal tech, this is where the code and contracts meet copyright law.
The industry has circled this fight before; this time the stakes feel different because legal precedent could change how every major model ingests and serves copyrighted text. Which side will blink first?