Anthropic Sued for Stealing Tens of Thousands of Songs to Train Claude

Seed story: "Anthropic sued over alleged theft of ‘tens of thousands’ of songs" (The Guardian) · search original Written from facts verified across 2 news report(s) — original explainer, not a copy or translation. Sources listed at the end.

The legal battle over AI training data just escalated with a new lawsuit alleging Anthropic scraped "tens of thousands" of songs from pirate sources to power its Claude models. For creators, this case is critical because it targets the removal of copyright-management information and seeks statutory damages up to $150,000 per work, potentially setting a precedent for how royalty claims are handled when AI companies use unlicensed music.

The Lawsuit: Allegations of Massive Copyright Infringement

On August 28, 2026, Sony Music Publishing and Warner Chappell Music filed a federal lawsuit in Northern California against Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann. The complaint alleges that Anthropic illegally downloaded, scraped, and torrented tens of thousands of copyrighted songs to train its Claude AI models. This marks the latest escalation in a dispute that began with earlier suits in October 2023 and January 2026.

The plaintiffs claim Anthropic sourced lyrics and sheet music from both pirate platforms and licensed services. Specifically, the complaint cites:

Pirate sources like Library Genesis and the Pirate Library Mirror
Licensed sites such as Musixmatch and LyricFind
Stripping of identifying information to deny attribution

Notable works named in the filing include Mariah Carey’s “All I Want for Christmas Is You” and Survivor’s “Eye of the Tiger.” Anthropic has stated it disagrees with the claims and intends to defend itself robustly in court.

From Pirate Libraries to AI Models

The complaint details a dual-sourcing strategy, alleging Anthropic harvested content from both illicit and licensed platforms. Specifically, the lawsuit claims the company downloaded lyrics and sheet music from pirate repositories like Library Genesis and the Pirate Library Mirror. Simultaneously, it allegedly scraped data from licensed services such as Musixmatch and LyricFind. This mixed approach reportedly allowed Anthropic to amass a vast dataset for training Claude, bypassing traditional licensing agreements.

Beyond the initial acquisition, the plaintiffs allege a deliberate effort to obscure ownership. During processing, Anthropic reportedly stripped copyright management information from the files. This action denied original creators attribution and made it difficult to track the provenance of the training data.

Pirate Sources: Library Genesis and the Pirate Library Mirror.
Licensed Sites: Musixmatch and LyricFind.
Data Manipulation: Removal of identifying metadata.

For creators, this highlights a critical risk: even if data originates from licensed sources, the subsequent stripping of metadata can sever the link between the work and its owner, complicating future royalty claims and contract enforcement.

The Financial Stakes: Statutory Damages and Precedent

The financial exposure in this case is staggering. Plaintiffs are seeking statutory damages of up to $150,000 for each infringed work, alongside $25,000 for every instance of stripped copyright-management information. Given the allegation that "tens of thousands" of songs were used, the potential liability could reach astronomical figures, far exceeding typical litigation costs.

This demand follows a significant precedent set in September 2025, when Anthropic settled a separate copyright dispute with US authors for $1.5 billion. That massive payout signals that the company is prepared to pay substantial sums to resolve legal threats.

Statutory Damages: Up to $150,000 per work.
Metadata Penalties: $25,000 per removed identifier.
Precedent: A prior $1.5 billion settlement with authors.

For creators, these figures suggest that AI companies may view litigation settlements as a manageable operational cost rather than a deterrent. This dynamic could influence future contract negotiations, as rights holders leverage the threat of high statutory damages to secure better terms or compensation for training data usage.

Why This Changes the Legal Landscape for Creators

Challenging the Fair Use Shield

This lawsuit directly tests the boundaries of the "fair use" defense, which many AI companies have relied upon to justify training on copyrighted material. By alleging that Anthropic sourced content from pirate libraries and stripped copyright-management information, the plaintiffs argue that the use was neither fair nor transformative. This approach moves the legal conversation beyond simple access to the integrity of the source, suggesting that laundering stolen data does not immunize a model from liability.

For creators, this sets a new benchmark for intellectual property enforcement. If the court rules against Anthropic, it could dismantle the blanket defense that AI training is inherently fair use.

Source Integrity: The case highlights that obtaining data from illicit sources weakens fair use claims.
Attribution Removal: Stripping metadata may be viewed as an intentional act to evade rights holders.
Precedent Setting: A ruling here could define the standard for all future AI training disputes.

Ultimately, this shifts the burden of proof, forcing companies to demonstrate clear, lawful acquisition of data to protect their models.

Implications for Creator Contracts and Rights

This litigation signals a critical shift in how AI companies approach data licensing. As Anthropic prepares for a stockmarket listing that could value it at $2 trillion, the risk of unlicensed training data becomes a material liability. Creators and publishers must now scrutinize existing contracts to ensure they explicitly address AI usage, moving beyond vague "digital rights" clauses to specific prohibitions on model training.

Key contractual adjustments may include:

Explicit AI Licensing: Clear terms defining whether data can be used for model training or fine-tuning.
Attribution Requirements: Mandating that copyright-management information remains intact, preventing the stripping of metadata alleged in this suit.
Revenue Sharing: Defining compensation structures for works used in generative outputs.

If courts uphold the plaintiffs' claims, artists may gain leverage to renegotiate deals, ensuring their intellectual property is not merely scraped but properly licensed and compensated in the AI era.

What Creators Should Do Now

Practical Steps for Protection

With the lawsuit alleging Anthropic stripped identifying information from songs, creators must actively monitor how their work appears in AI outputs. Regularly search for your titles and lyrics within Claude to detect unauthorized usage. If you find your content, document the date, prompt, and output immediately. This evidence is critical for proving infringement and supporting potential royalty claims.

To strengthen your legal position, ensure your contracts explicitly address AI training and data scraping. Consider adding clauses that require attribution and prohibit the use of your work in model development without consent.

Audit your portfolio: Identify high-value tracks most likely to be scraped.
Monitor AI platforms: Use automated tools or manual checks to track appearances.
Update contracts: Include clear AI usage restrictions and attribution requirements.
Preserve evidence: Save screenshots and metadata of any unauthorized AI outputs.

FAQ

What specific allegations does the new lawsuit make against Anthropic regarding song theft?

The lawsuit filed by Sony Music Publishing and Warner Chappell Music alleges that Anthropic illegally downloaded, scraped, and torrented tens of thousands of copyrighted songs to train its Claude AI models. The complaint claims the company obtained lyrics and sheet music from pirate sources like Library Genesis as well as licensed sites, while also stripping identifying information to deny copyright owners attribution.

How much financial damages are the plaintiffs seeking in this copyright case?

The plaintiffs are seeking statutory damages of up to $150,000 for each infringed work. Additionally, they are requesting $25,000 for every instance where copyright-management information was allegedly removed during the processing of the songs.

What is Anthropic's official response to the accusations of stealing music?

Anthropic has stated that it disagrees with the publishers' claims and intends to defend itself robustly in court. This legal challenge comes as the company prepares for a stockmarket listing that could value it at $2 trillion, following a separate $1.5 billion settlement with US authors in September 2025.