Concord Music Group, Universal Music Publishing Group and ABKCO Music have filed an amended complaint against Anthropic over the alleged copying of copyrighted song lyrics by its Claude chatbot.
The publishers filed the Second Amended Complaint on Wednesday (July 22) in the US District Court for the Northern District of California — an expanded version of the same amended complaint they first put forward in May.
It updates the case they first brought in October 2023, which covers around 500 songs and is the first of two separate copyright lawsuits the group has filed against Anthropic.
Their second, separate suit, filed in January 2026, covers more than 20,000 songs and seeks over $3 billion over Anthropic‘s alleged torrenting of lyrics from pirate “shadow” libraries.
The amended complaint also lands as Anthropic works through a separate $1.5 billion settlement it agreed in September 2025 with a group of authors over its use of pirated books to train Claude.
The 71-page pleading draws heavily on Anthropic‘s own internal records, obtained through discovery, to support the infringement claims.
It seeks to hold Anthropic responsible for what the complaint calls the “systematic and widespread infringement of their copyrighted song lyrics” through the training and operation of its Claude AI models.
The filing states that Anthropic co-founder and chief compute officer Tom Brown once queried “@Claude what are the lyrics to desolation row by Dylan?” in an internal record.
The query references the Bob Dylan composition Desolation Row, one of the works in the suit.
The complaint further alleges that Anthropic and the temporary workers it hired to “finetune” its models through reinforcement learning prompted Claude with requests for the publishers’ lyrics during development.
The publishers argue this shows Anthropic “intended, expected, and designed its AI models to respond to prompts for Publishers’ and others’ lyrics and to copy the lyrics in output – as a feature, not a bug.”
The complaint also alleges that Anthropic used extraction tools to strip copyright notices and other identifying details from lyrics during training, dismissing that information as “useless junk” in internal conversations.
It claims co-founders Ben Mann and Jared Kaplan debated which extraction tools to use in June 2021, ultimately choosing one that removed copyright notices entirely.
The filing further alleges that the term “lyric” appeared in more than 170,000 Claude prompt and output records during a nine-day window in September 2023, the month before the suit was filed.
The publishers allege that Claude reproduces their lyrics both as training input and in its output to users, and that it does so even when users do not ask for lyrics.
The complaint says that when investigators asked Claude for the words to songs including Roar, I Will Survive, American Pie and What a Wonderful World, the model returned identical or near-identical copies.
It adds that Claude also produced lyrics in response to prompts that named no song at all.
Asked to “write a song about the death of Buddy Holly,” for example, the model returned the words to American Pie.
“ANTHROPIC’S COPYRIGHT INFRINGEMENT IS NOT INNOVATION; IN LAYMAN’S TERMS, IT’S THEFT.”
CONCORD, UMPG AND ABKCO’S AMENDED COMPLAINT
The publishers say they recognize the potential of ethical AI, but argue that Anthropic‘s conduct crosses a line.
“By contrast, Anthropic‘s copyright infringement is not innovation,” the complaint reads. “In layman’s terms, it’s theft.”
The publishers are seeking statutory damages of up to $150,000 per work.
In January 2025, the court approved a stipulation requiring Anthropic to maintain “guardrails” meant to stop Claude from reproducing copyrighted lyrics in its outputs.
In March 2025, Judge Eumi K. Lee denied the publishers’ request for a preliminary injunction that would have barred Anthropic from training on their lyrics, finding any harm could be compensated with money damages.
The same set of orders granted the publishers discovery, requiring Anthropic to hand over a sample of Claude prompt and output records – the material now cited throughout the amended complaint.
The publishers have since narrowed the case against Anthropic to direct infringement and the alleged removal of copyright management information, dropping earlier contributory and vicarious infringement claims.
They moved for partial summary judgment in March 2026, asking the court to reject Anthropic‘s fair use defense ahead of trial.
Anthropic is contesting the case and has argued that training Claude on copyrighted material amounts to fair use.
Anthropic had not publicly responded to the amended complaint at the time of writing.
Anthropic‘s valuation has risen since the case began.
The company confidentially filed for an initial public offering in June 2026, days after a $65 billion funding round valued it at $965 billion.
That valuation lifted Anthropic above rival OpenAI for the first time.
Anthropic said its run-rate revenue crossed $47 billion in May.
The amended complaint is one of several actions the music industry has brought against Anthropic.
BMG filed a third suit in March 2026, citing 493 compositions and alleging the AI firm had built its business on unlicensed works.
Elsewhere in AI litigation, Sony Music filed a new lawsuit against AI music platform Udio on Monday (July 20).
Udio’s competitor Suno faces expanding infringement claims from Universal and Sony.Music Business Worldwide
Concord Music Group, Universal Music Publishing Group, and ABKCO Music have escalated their legal battle against Anthropic, the creator of the Claude chatbot, by filing a Second Amended Complaint in the US District Court for the Northern District of California. This legal action, submitted on July 22, 2026, updates an earlier complaint filed in October 2023, which accused Anthropic of infringing on the copyrights of approximately 500 songs. This lawsuit is one of two copyright cases initiated by the publishers against Anthropic, with the second case—filed in January 2026—targeting over 20,000 songs and seeking more than $3 billion in damages for the alleged misuse of lyrics sourced from pirate « shadow » libraries.
The amended complaint is significant, especially in light of Anthropic’s recent $1.5 billion settlement related to another lawsuit concerning the use of pirated books to train Claude, which it agreed to in September 2025. The new 71-page complaint extensively references internal records from Anthropic, obtained through discovery, to substantiate claims of infringement. It asserts that Anthropic has systematically and widely infringed upon the publishers’ copyrighted lyrics during the training and operation of the Claude AI model.
One of the critical pieces of evidence mentioned in the complaint is a query made by Anthropic co-founder Tom Brown, which asked for the lyrics to Bob Dylan’s « Desolation Row. » This particular query is cited as indicative of Anthropic’s intent to incorporate copyrighted lyrics into its AI model’s capabilities. Furthermore, the complaint alleges that Anthropic, along with temporary workers hired to refine Claude through reinforcement learning, actively prompted the AI with requests for song lyrics from the publishers during its development phase. The publishers argue that this indicates a design choice by Anthropic to enable Claude to respond to prompts with copyrighted lyrics, presenting this as a deliberate feature rather than an unintended flaw.
The complaint also claims that Anthropic utilized various extraction tools to strip copyright notices and other identifying information from lyrics, deeming such data as « useless junk. » This was allegedly part of a systematic effort to remove copyright protection from the lyrics used to train its models. The filing indicates that in a nine-day window in September 2023, the term « lyric » appeared in over 170,000 Claude prompt and output records, suggesting extensive interaction with the publishers’ copyrighted material.
The complaint outlines specific instances where Claude reproduced lyrics verbatim, even when users did not explicitly request them. For example, Claude provided lyrics to well-known songs like « Roar, » « I Will Survive, » « American Pie, » and « What a Wonderful World » in response to prompts. In one notable instance, when asked to write a song about Buddy Holly’s death, Claude returned the lyrics to « American Pie, » highlighting the AI’s tendency to generate copyrighted content without direct prompts.
The publishers acknowledge the potential benefits of ethical AI but assert that Anthropic’s actions cross a legal and ethical line. They firmly state in the complaint that Anthropic’s copyright infringement should not be misconstrued as innovation, labeling it as « theft. » As part of their legal strategy, the publishers are seeking statutory damages that could reach up to $150,000 for each work infringed.
Previously, in January 2025, the court mandated that Anthropic implement « guardrails » to prevent Claude from reproducing copyrighted lyrics. However, a preliminary injunction that would have barred Anthropic from training on the publishers’ lyrics was denied in March 2025, with the court concluding that any potential harm could be addressed through monetary damages. The subsequent court orders allowed the publishers to conduct discovery, which has played a significant role in shaping the allegations in the amended complaint.
As the case evolves, the publishers have narrowed their focus to direct infringement and the alleged removal of copyright management information, previously dropping claims related to contributory and vicarious infringement. They have also moved for partial summary judgment, asking the court to dismiss Anthropic’s fair use defense before the trial commences. In response, Anthropic has maintained that its training practices fall under fair use, but the specifics of their defense have not been publicly articulated at the time of the amended complaint’s filing.
Since the initiation of the legal proceedings, Anthropic’s valuation has significantly increased. The company confidentially filed for an initial public offering in June 2026, shortly after securing a $65 billion funding round that valued it at $965 billion, surpassing rival OpenAI for the first time. As of May 2026, Anthropic reported an annual revenue run-rate exceeding $47 billion.
The legal challenges facing Anthropic are not limited to this particular case. The music industry has been increasingly vigilant in addressing copyright issues related to AI. In a separate action, BMG filed a lawsuit in March 2026, claiming that Anthropic built its business on unlicensed works, citing 493 compositions. In addition, Sony Music has also initiated legal proceedings against the AI music platform Udio, and its competitor, Suno, is facing growing infringement claims from Universal and Sony.
In summary, the ongoing legal battles highlight the tensions between the burgeoning field of AI development and the established rights of content creators. The outcome of these cases may set crucial precedents in the intersection of copyright law and artificial intelligence.
