Sony Music, Warner sue Anthropic, alleging a “brazen campaign” of intellectual property theft

Last updated: September 13, 2026

Sony Music Entertainment and Warner Music Group have filed a sweeping copyright infringement lawsuit against AI startup Anthropic, marking a significant escalation in the music industry’s battle to control how its intellectual property is used to train generative models. The complaint, filed in a Tennessee federal court, accuses the Amazon-backed company of a “brazen campaign” of illegal piracy, alleging that Anthropic systematically copied and ingested vast catalogs of song lyrics to build its Claude chatbot without permission or compensation.

Harvard’s $699 startup bootcamp offers AI avatars of its…

Background

The legal action represents the most aggressive move yet by major record labels against an AI foundation model provider. While previous disputes have often centered on the output of AI tools—such as voice clones or generated melodies—this complaint focuses squarely on the input side: the training data itself. The plaintiffs allege that Anthropic scraped lyrics from countless websites, effectively building a commercial product on the back of copyrighted text owned by publishers and songwriters.

The suit contends that when users prompt Claude to reproduce lyrics—whether for a specific hit song or a request like “write a song in the style of [Artist Name]”—the model frequently generates near-verbatim copies of protected works. The labels argue this constitutes direct copyright infringement and vicarious liability, asserting that Anthropic has the technical capability to filter or block such outputs but has failed to implement adequate guardrails.

Anthropic, founded by former OpenAI researchers and valued at over $18 billion, has positioned itself as a safety-focused alternative in the AI race. The company has previously argued that its training processes constitute fair use and that its models transform source material rather than merely memorizing it. This lawsuit tests those defenses in a jurisdiction known for favorable precedent toward rights holders.

Why It Matters

For businesses integrating AI into their workflows, this case highlights a critical and often overlooked risk vector: upstream liability. Most enterprises do not train their own foundation models; they license APIs from providers like Anthropic, OpenAI, or Google. If courts rule that the underlying training data was illegally obtained, the legal exposure could theoretically extend downstream to companies deploying those models in customer-facing products, marketing content, or internal tools.

The music industry’s strategy signals a broader shift. Rights holders are no longer waiting for regulatory clarity; they are using litigation to force licensing negotiations. We have already seen this play out in the visual arts and publishing sectors, where companies like Adobe and OpenAI have struck deals with stock photo agencies and news publishers respectively. The music sector, historically litigious and highly organized, is now applying maximum pressure to establish a market standard: pay for training data, or face the courts.

For AI adopters, the immediate takeaway is due diligence. Procurement and legal teams need to ask vendors pointed questions about training data provenance, indemnification clauses, and the status of ongoing litigation. A model that performs well today may become a liability tomorrow if its legal foundation crumbles.

What’s Next

The case is likely to be a marathon, not a sprint. Anthropic will almost certainly move to dismiss on fair use grounds, arguing that training on publicly available text is transformative and non-expressive. The courts will eventually have to define the boundaries of “fair use” in the context of large language models—a question the Supreme Court has so far avoided.

In the meantime, expect a flurry of licensing announcements. Anthropic and its competitors have a strong financial incentive to settle with major rights holders to remove the cloud of uncertainty hanging over their enterprise sales cycles. We may see the emergence of “clean” model tiers—versions trained exclusively on licensed or public domain data—offered at a premium to risk-averse corporate clients.

Legislative action remains a wild card. The NO FAKES Act and similar bills in Congress aim to create federal property rights in voice and likeness, while the EU AI Act imposes transparency requirements for training data. However, US copyright law moves at the speed of litigation, not legislation.

The Bottom Line

Business owners should treat this lawsuit as a wake-up call, not a spectator sport. The era of “move fast and break things” is colliding with centuries of IP law. Until the dust settles, the safest path is to demand transparency from your AI vendors, secure contractual indemnification, and avoid building mission-critical workflows on models with unresolved, high-stakes copyright exposure. The cost of compliance is rising; the cost of ignorance could be existential.

Source: Original Article

For more on this, see Final 24 hours to save up to $410 on your TechCrunch Disrupt 2026 ticket.

Related: Anthropic CEO outlines plan to ‘pace the frontier’, Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data, Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too.

Share this article