Blab Music
Music Publishers Refile Direct Infringement Claims Against Anthropic Amid Supreme Court Shift
BMG Rights Management and Concord Music Group filed a second amended complaint on July 22 2026, reigniting a lawsuit against AI developer Anthropic. The new pleading zeroes in on direct copyright infringement and the deliberate removal of copyright‑management information (CMI) from the data that trains Claude, the company’s flagship language model.
The filing comes after the Supreme Court’s March 25 decision in Cox Communications, Inc. v. Sony Music Entertainment, which narrowed the scope of secondary liability for internet service providers. By trimming the legal safety net that had previously supported contributory and vicarious claims, the Court has forced rights holders to rethink their approach. In this case, publishers are betting that a direct‑infringement strategy will survive the tightened legal framework.
According to the complaint, Anthropic copied lyrics from publicly accessible websites that contained copyrighted text and then stripped the accompanying CMI before incorporating the material into its training set. The suit further alleges that the same extraction tools were used to clean CMI from third‑party datasets such as Common Crawl and the large‑scale text collection known as The Pile.
Earlier filings had seen the publishers withdraw their contributory and vicarious claims in a second amended complaint. However, a court‑issued redaction order removed certain details, leaving only a single sentence that references the alleged use of unauthorized training data and CMI removal. The redacted portion, not publicly available, is said to describe how Anthropic “deliberately used these same datasets—including The Pile—that it knew had been systematically stripped of copyright management information.”
The Concord lawsuit, originally filed in 2024 and refiled in 2026, also accuses Anthropic’s co‑founders Dario Amodei and Benjamin Mann of contributory infringement through torrenting. In addition to direct infringement claims, the complaint asserts that Anthropic’s training and output activities violate copyright law.
A judge has scheduled a case‑management conference for September 2 to chart the next steps for both cases. The meeting will address procedural matters and the parties’ positions on discovery and potential settlement.
This litigation is part of a broader wave of copyright disputes involving AI developers. In March 2026, Anthropic settled the author class action Bartz v. Anthropic for $1.5 billion. The company is also facing a motion for partial summary judgment from Universal Music Publishing Group, Concord, and ABKCO in the Northern District of California.
Founded in 2021 by former OpenAI employees, Anthropic builds its Claude models on large‑scale text corpora. Its privacy policy states that it may use user conversations for training if flagged for safety review, but it does not disclose the exact volume of data used. The redactions in the current complaint reflect the company’s efforts to keep training‑data figures confidential.
The Supreme Court’s ruling in Cox has implications for secondary liability claims. By reversing a Fourth Circuit decision that had held an internet service provider liable for users’ infringement, the Court has tightened the legal standard for contributory and vicarious claims. The publishers’ focus on direct infringement reflects a strategic shift in response to the Court’s narrowing of secondary liability.
The outcome of the July filings will shape how music publishers pursue enforcement against AI systems that rely on copyrighted text. The case‑management conference in September is likely to set the pace for discovery and potential settlement negotiations.
Until the court’s next ruling, the lawsuits remain pending. Anthropic has not issued a public statement regarding the new complaint, and the publishers have not released additional details beyond the filed documents.
The legal battles underscore the growing tension between AI developers and rights holders over the use of copyrighted material in training large language models. The proceedings will be closely watched by industry stakeholders as they navigate the evolving intersection of technology and intellectual property law.
The filing comes after the Supreme Court’s March 25 decision in Cox Communications, Inc. v. Sony Music Entertainment, which narrowed the scope of secondary liability for internet service providers. By trimming the legal safety net that had previously supported contributory and vicarious claims, the Court has forced rights holders to rethink their approach. In this case, publishers are betting that a direct‑infringement strategy will survive the tightened legal framework.
According to the complaint, Anthropic copied lyrics from publicly accessible websites that contained copyrighted text and then stripped the accompanying CMI before incorporating the material into its training set. The suit further alleges that the same extraction tools were used to clean CMI from third‑party datasets such as Common Crawl and the large‑scale text collection known as The Pile.
Earlier filings had seen the publishers withdraw their contributory and vicarious claims in a second amended complaint. However, a court‑issued redaction order removed certain details, leaving only a single sentence that references the alleged use of unauthorized training data and CMI removal. The redacted portion, not publicly available, is said to describe how Anthropic “deliberately used these same datasets—including The Pile—that it knew had been systematically stripped of copyright management information.”
The Concord lawsuit, originally filed in 2024 and refiled in 2026, also accuses Anthropic’s co‑founders Dario Amodei and Benjamin Mann of contributory infringement through torrenting. In addition to direct infringement claims, the complaint asserts that Anthropic’s training and output activities violate copyright law.
A judge has scheduled a case‑management conference for September 2 to chart the next steps for both cases. The meeting will address procedural matters and the parties’ positions on discovery and potential settlement.
This litigation is part of a broader wave of copyright disputes involving AI developers. In March 2026, Anthropic settled the author class action Bartz v. Anthropic for $1.5 billion. The company is also facing a motion for partial summary judgment from Universal Music Publishing Group, Concord, and ABKCO in the Northern District of California.
Founded in 2021 by former OpenAI employees, Anthropic builds its Claude models on large‑scale text corpora. Its privacy policy states that it may use user conversations for training if flagged for safety review, but it does not disclose the exact volume of data used. The redactions in the current complaint reflect the company’s efforts to keep training‑data figures confidential.
The Supreme Court’s ruling in Cox has implications for secondary liability claims. By reversing a Fourth Circuit decision that had held an internet service provider liable for users’ infringement, the Court has tightened the legal standard for contributory and vicarious claims. The publishers’ focus on direct infringement reflects a strategic shift in response to the Court’s narrowing of secondary liability.
The outcome of the July filings will shape how music publishers pursue enforcement against AI systems that rely on copyrighted text. The case‑management conference in September is likely to set the pace for discovery and potential settlement negotiations.
Until the court’s next ruling, the lawsuits remain pending. Anthropic has not issued a public statement regarding the new complaint, and the publishers have not released additional details beyond the filed documents.
The legal battles underscore the growing tension between AI developers and rights holders over the use of copyrighted material in training large language models. The proceedings will be closely watched by industry stakeholders as they navigate the evolving intersection of technology and intellectual property law.
