Anthropic Sued Over Alleged Theft of Tens of Thousands of Songs

Seed story: "Anthropic sued over alleged theft of ‘tens of thousands’ of songs | Music industry" (The Guardian) · search original Written from facts verified across 3 news report(s) — original explainer, not a copy or translation. Sources listed at the end.

With Sony Music Publishing and Warner Chappell Music alleging that Anthropic scraped “tens of thousands” of copyrighted songs from pirate sources to train its Claude models, the latest lawsuit signals a sharp escalation in the fight over AI training data. As the company faces statutory damages claims of up to $150,000 per infringed work and navigates its fifth music-copyright suit, creators must now reckon with how these legal battles and prior settlements are reshaping the landscape of ownership and compensation for machine learning.

The Lawsuit: Allegations of Mass Scraping and Piracy

At the core of the complaint is the allegation that Anthropic bypassed legal licensing channels to build its dataset. The plaintiffs claim the company illegally downloaded, scraped, and torrented thousands of copyrighted songs to train its Claude models. This approach allegedly relied on a mix of illicit and legitimate sources:

Pirate repositories such as Library Genesis and the Pirate Library Mirror
Licensed data providers like Musixmatch and LyricFind

By sourcing lyrics and sheet music from these platforms, the lawsuit argues that Anthropic engaged in mass piracy rather than authorized data acquisition. This specific mechanism of obtaining training data is central to the claim of willful infringement.

For creators, this distinction matters significantly. If a model is trained on pirated content, it may undermine arguments that the AI company had a good-faith belief in the legality of its data usage. Such findings could weaken fair use defenses and strengthen claims for statutory damages, directly impacting how rights holders negotiate future licensing agreements with AI developers.

Why This Case Differs from Previous AI Litigation

This lawsuit marks a significant escalation because it alleges intentional piracy rather than accidental scraping. Unlike prior disputes where companies claimed they merely accessed public data, the complaint specifically identifies Anthropic’s use of illicit platforms like Library Genesis and the Pirate Library Mirror. This distinction is critical, as it suggests a deliberate strategy to bypass licensing fees and copyright protections.

Intentional Piracy: Alleged use of specific pirate sites to obtain lyrics and sheet music.
Targeted Sources: Claims of downloading from both illegal mirrors and licensed sites like Musixmatch.
Legal Precedent: Contrasts with earlier cases that focused on general web scraping.

For creators, this shift from "unauthorized access" to "willful infringement" has profound implications. If courts accept these allegations, it could establish a stricter liability standard for AI developers. This may force companies to implement more rigorous provenance tracking, potentially altering how training data is sourced and how creators are compensated for their works in future contracts.

The Financial Stakes: Statutory Damages and CMI Removal

The financial exposure in this case is staggering due to the specific statutory damages sought by the plaintiffs. Sony Music Publishing and Warner Chappell Music are pursuing up to $150,000 for each willfully infringed song. Given the complaint alleges the theft of "tens of thousands" of tracks, the cumulative liability could reach billions of dollars. This figure dwarfs the $1.5 billion settlement Anthropic reportedly reached in the Bartz case in September 2025, signaling a significant escalation in the cost of AI training data disputes.

Beyond direct infringement, the lawsuit targets the removal of copyright management information (CMI). Plaintiffs are seeking an additional $25,000 for every instance where such metadata was stripped. This dual-pronged approach creates a compounding financial risk that extends beyond simple licensing fees.

$150,000 per song for willful infringement
$25,000 per instance of CMI removal
$1.5 billion in the prior Bartz settlement

For creators, this highlights that AI companies face severe penalties for both using content and obscuring its provenance, potentially influencing future contract negotiations regarding data attribution and royalty structures.

Anthropic’s Defense and the Broader Legal Landscape

Anthropic has firmly rejected the allegations, stating it intends to defend itself robustly in court. This stance places the new filing within a rapidly expanding legal landscape. According to reports, Anthropic is currently facing five active music-copyright suits. These include previous cases involving Universal Music Publishing, Concord Music Group, ABKCO, and BMG, alongside the new claims from Sony Music Publishing and Warner Chappell Music.

The sheer volume of litigation underscores the industry’s aggressive stance against AI training data practices. Notably, Anthropic recently settled the Bartz v. Anthropic case in September 2025 for $1.5 billion. This massive payout signals that the company is prepared to pay substantial sums to resolve disputes. For creators, this context is critical. It suggests that while individual lawsuits may be defended vigorously, the cumulative financial pressure from multiple active cases can force significant settlements. Consequently, the legal environment is shifting from isolated disputes to a systemic challenge regarding how AI companies acquire and use copyrighted works for model development.

Implications for Creator Contracts and Rights

Shifting Fair Use Boundaries

This litigation challenges the prevailing assumption that training AI models constitutes fair use. By alleging willful piracy rather than mere access, the plaintiffs aim to redefine the legal threshold for machine learning. If the court rejects Anthropic’s defense, it may establish that scraping copyrighted works for commercial training is not protected, fundamentally altering the legal landscape for generative AI.

For creators, this shift has immediate contractual implications:

Licensing Clarity: Contracts may need explicit clauses defining data usage rights, moving beyond vague "training data" permissions.
Rights Retention: Creators must ensure their agreements prevent unauthorized scraping or torrenting of their work.
Payment Structures: Future deals may require separate compensation for AI training, distinct from traditional performance or mechanical royalties.

As the legal definition tightens, creators should review existing contracts to verify that their rights are protected against unauthorized AI ingestion.

What Creators Should Do Now

Proactive Protection Strategies

With five active music-copyright suits against Anthropic, including recent filings by Sony Music Publishing and Warner Chappell, creators must assume their work is a target. The complaint alleges scraping from both pirate sources like Library Genesis and licensed platforms such as Musixmatch. This dual-source approach means that even legitimate licensing does not guarantee safety from unauthorized AI training use.

To secure intellectual property, rights holders should adopt a multi-layered monitoring strategy:

Audit Licensing Agreements: Review contracts to ensure they explicitly prohibit AI training and data scraping.
Monitor Pirate Sites: Regularly check known repositories like the Pirate Library Mirror for unauthorized uploads of your lyrics or sheet music.
Track Licensed Platforms: Verify how data is handled on services like LyricFind to detect potential misuse.

Acting now is critical. As seen in the Bartz settlement, early legal action can lead to significant financial recovery. By documenting unauthorized usage and understanding statutory damages of up to $150,000 per song, creators can better protect their rights and negotiate stronger terms in future contracts.

FAQ

What specific actions are Anthropic accused of in the new lawsuit?

Sony Music Publishing and Warner Chappell Music allege that Anthropic illegally downloaded, scraped, and torrented thousands of copyrighted songs to train its Claude AI models. The complaint claims these materials were obtained through pirate sources like Library Genesis and licensed sites such as Musixmatch and LyricFind.

How much damages are the plaintiffs seeking in this case?

The lawsuit seeks statutory damages of up to $150,000 for each willfully infringed song and up to $25,000 for each instance of removed copyright-management information. These figures are part of the claim filed in the U.S. District Court for the Northern District of California.

How does Anthropic respond to these copyright allegations?

Anthropic has rejected the allegations and stated that it intends to defend itself robustly in court. This is the fifth music-copyright suit against the company, following a previous $1.5 billion settlement in the Bartz v. Anthropic case in September 2025.