Sony and Warner Sue Anthropic: The $2 Trillion Copyright Reckoning That Could Reshape AI Training

On August 28, 2026, Sony Music Publishing and Warner Chappell Music filed a federal lawsuit against Anthropic that could reshape how AI companies acquire training data. The complaint, filed in the US District Court for the Northern District of California, alleges “one of the largest and most blatant ongoing thefts of intellectual property in history.”

The stakes are staggering. The complaint identifies “tens of thousands” of copyrighted musical compositions. At $150,000 per willfully infringed work, Anthropic’s theoretical statutory exposure is in the multi-billion-dollar range. And this is not an isolated case — it’s the fourth copyright lawsuit filed against Anthropic by music publishers alone in 2026.

The Lawsuit: What’s Alleged

The 48-page complaint names three defendants: Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann. The allegations are specific and damaging:

The Songs

The complaint lists copyrighted works including “Ain’t No Mountain High Enough” (Marvin Gaye and Tammi Terrell), “All I Want for Christmas Is You” (Mariah Carey), “Eye of the Tiger,” “Here Comes Santa Claus,” “Paper Rings” (Taylor Swift), and songs by The Beatles, Michael Jackson, and hundreds of other artists.

The Method

The complaint alleges Anthropic illegally obtained the publishers’ lyrics and sheet music through torrent downloads. It identifies specific instances:

• June 2021: Co-founder Benjamin Mann used BitTorrent to download at least 5 million pirated books from Library Genesis
• July 2022: Anthropic employees torrented at least 2 million more from Pirate Library Mirror
• Anthropic also scraped lyrics from licensed sites including MusixMatch and LyricFind
• Anthropic ran a “destructive scanning” operation on second-hand books

The Output

The complaint alleges Claude can reproduce copyrighted lyrics “verbatim” when prompted, and that Anthropic used the lyrics to train Claude to “generate vast quantities of purportedly ‘new’ AI-generated song lyrics, which compete with Music Publishers’ legitimate copyrighted works as harmful market substitutes.”

The Legal Claims

Four counts:

1. Direct infringement by torrenting — against all three defendants
2. Contributory infringement by torrenting — against Amodei and Mann personally
3. Direct infringement — against Anthropic
4. Removal or alteration of copyright management information — against Anthropic

Sony and Warner are demanding a jury trial, statutory damages of up to $150,000 per willfully infringed work, up to $25,000 per violation for removing copyright management information, destruction of all infringing copies, and an accounting of Claude’s training data.

The Bartz Precedent: Fair Use vs. Piracy

The legal landscape was shaped by the Bartz v. Anthropic case, where Anthropic became the first AI company to settle a copyright class action, paying $1.5 billion to a group of authors in 2025.

The Bartz ruling established a critical distinction:

• Training generative AI models on copyrighted content IS transformative fair use — the use is sufficiently transformative to qualify for fair use protection
• But acquiring that content through piracy is NOT legal — the method of acquisition matters as much as the use

This distinction is the crux of the Sony/Warner case. Anthropic’s defense — “training generative AI models is a transformative fair use, as the court held in Bartz” — may hold for the training claim. But the complaint alleges specific instances of illegal torrenting, including by the co-founder personally. If those allegations are proven, fair use may not apply to the acquisition.

Sony and Warner made this argument explicitly in their complaint: “Anthropic clearly considers that to be just the cost of doing business given that its entire business model continues to be built on copyright theft. And $1.5bn is obviously not a large enough settlement to deter infringing conduct by a company that has parlayed such mass infringement into a staggering 2-trillion-dollar valuation.”

The Legal Wave: All Three Major Publishers Now Litigating

The Sony/Warner lawsuit is not an isolated case. It’s part of a legal wave:

Plaintiff	Date	Scope	Damages Sought
Universal Music + Concord	January 2026	20,000+ works	$3 billion+
BMG	March 2026	493 compositions	Unspecified
Round Hill Music	August 17, 2026	Unspecified	Unspecified
Sony + Warner	August 28, 2026	“Tens of thousands”	$150K per work

All three major music publisher groups — Universal, Sony, and Warner — are now litigating against Anthropic. The same law firm, Oppenheim + Zebrak, represents the publishers in both the Concord/UMG case and the Sony/Warner case.

This is not one lawsuit. It’s a coordinated legal strategy.

Anthropic’s Defense

An Anthropic spokesperson said: “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court.” The company calls the lawsuit “recycled” and references the Bartz ruling: “Training generative AI models is a transformative fair use — as the court held in Bartz.”

The challenge: the complaint alleges specific, documented instances of illegal torrenting by named individuals, including the co-founder. Fair use protects the act of training, but it doesn’t protect the act of piracy. If the torrenting allegations are proven, Anthropic’s fair use defense may not cover the acquisition of the content.

The personal allegations against Benjamin Mann are particularly significant. The complaint alleges he used BitTorrent to download 5 million pirated books from Library Genesis in June 2021 — before Anthropic was publicly known. This predates many of the company’s public commitments to responsible AI development.

What This Means for Business Leaders

1. Audit Your AI Vendors’ Training Data

If your organization uses AI models from Anthropic, OpenAI, Google, or any other lab, you need to understand where their training data came from. The legal risk of using models trained on pirated content is not yet settled — but the direction is clear.

Ask your vendors:

• What are your training data sources?
• Do you have licenses for copyrighted content?
• What is your policy on pirated content?
• Do you indemnify customers against IP claims?

The AI Strategy for Business consultation now includes an AI Vendor Risk Assessment: evaluating your AI vendors’ training data practices, contractual protections, and legal exposure.

2. Ensure Contracts Include IP Indemnification

Your contracts with AI vendors should include indemnification clauses that protect you against IP infringement claims. If a vendor’s model was trained on pirated content and that content’s owner sues, you need contractual protection.

Key clauses to include:

• Indemnification for IP claims — the vendor covers your legal costs
• Warranties about training data — the vendor warrants that training data was legally acquired
• Right to audit — you can inspect the vendor’s training data practices
• Termination rights — you can terminate if the vendor is found to have used pirated content

3. Build Documented, Licensed Data Pipelines

If your organization is building or fine-tuning its own AI models, the Bartz ruling provides a clear framework: training on copyrighted content may be fair use, but acquiring content through piracy is not.

This means your data pipelines need to be:

• Documented — every data source recorded with acquisition method
• Licensed — content acquired through legitimate channels (APIs, licenses, public domain)
• Auditable — you can demonstrate how each dataset was acquired
• Clean — no pirated content, no torrent downloads, no scraping behind paywalls

The CTO Technology Advisory service helps organizations build compliant data pipelines: licensed data sources, documentation frameworks, and audit trails that withstand legal scrutiny.

4. Monitor AI Outputs for Verbatim Reproduction

The Sony/Warner complaint alleges Claude can reproduce copyrighted lyrics “verbatim” when prompted. If your organization’s AI outputs include verbatim reproduction of copyrighted content, you face legal risk — even if the training itself is fair use.

Implement:

• Output filters that detect and block verbatim reproduction of known copyrighted content
• Prompt monitoring that flags requests likely to produce infringing outputs
• Usage policies that prohibit employees from using AI to reproduce copyrighted content

5. Track the Legal Landscape

The AI copyright landscape is evolving rapidly. Key developments to watch:

• The Sony/Warner case — if it goes to trial, the discovery process could reveal Anthropic’s full training data practices
• The Universal/Concord case — seeking $3B+ for 20,000+ works
• Potential regulatory action — the EU AI Act includes provisions on training data transparency
• Settlements vs. verdicts — each settlement normalizes the cost of piracy as a business expense; each verdict establishes legal precedent

The Executive AI Workshop includes sessions on AI legal risk: helping leadership teams understand the copyright landscape, audit their AI vendors, and build compliance frameworks.

The Bottom Line

The Sony/Warner lawsuit is not just another copyright case. It’s a test of whether the AI industry’s training data practices are sustainable. The Bartz ruling established that fair use protects training but not piracy. The Sony/Warner case will test whether that distinction holds when the piracy is alleged to have been committed by the company’s co-founder personally.

For business leaders, the message is clear: the era of “train first, ask permission later” is ending. Organizations that use AI need to understand where their models’ training data came from, ensure their contracts protect them, and build their own data pipelines on licensed, documented foundations.

The $1.5 billion Bartz settlement didn’t deter Anthropic. Sony and Warner are betting that multi-billion-dollar statutory damages will. Whether it does or not, the legal landscape has shifted — and organizations that ignore the shift do so at their own peril.

The copyright reckoning is here. The question is whether your organization is prepared for it.