Sony and Warner Sue Anthropic for Allegedly Using Tens of Thousands of Copyrighted Songs to Train Claude
Two of the world's biggest music publishers say the company behind the Claude chatbot trained its AI on their songwriters' work without permission. The lawsuit could reach billions of dollars.

Key points
- Sony Music Publishing and Warner Chappell filed a multi-billion-dollar lawsuit against Anthropic, the AI company behind the Claude chatbot.
- The publishers claim Anthropic used "tens of thousands" of copyrighted songs to train Claude without paying rights holders.
- Sony Music Publishing and Warner Chappell manage copyright on behalf of songwriters and composers, not record labels.
- No settlement figure has been disclosed; the case is ongoing.
Two music giants are taking Anthropic to court, claiming the San Francisco AI startup fed their songwriters' work into its computers without asking or paying.
Sony Music Publishing and Warner Chappell, the arms of those companies that manage song rights on behalf of the people who actually write the music, say Anthropic used "tens of thousands" of copyrighted works to train Claude. Claude is a large language model, the same type of AI technology that powers ChatGPT, which can hold conversations, write text, and answer questions. The lawsuit seeks damages that could reach into the billions, as first reported by The Guardian.
What exactly is the complaint?
The publishers say Anthropic copied protected song lyrics and compositions to build a commercial product and never compensated the creators.
Music publishers like Sony Music Publishing and Warner Chappell do not own the recordings you hear on streaming services. They own the underlying compositions: the words and the melody as written on paper. When a song appears in a film, or an AI model trains on its lyrics, those publishers are the ones who need to grant permission and collect fees on the songwriter's behalf.
The allegation is that Anthropic skipped that step entirely.
Does Anthropic have a defence?
Anthropics has not yet filed its formal response, so its legal argument is not yet on the record.
AI companies have generally argued that training on publicly available text and lyrics falls under "fair use", a legal doctrine that allows limited reproduction of copyrighted material without permission in certain circumstances. Courts in the United States are still working out whether mass AI training qualifies. Several similar cases involving book authors, visual artists, and news publishers are moving through the system at the same time, so the outcome of any one of them could shape the rest.
What does this mean for the people who wrote the songs?
For now, nothing changes day to day for songwriters. But the case matters to them.
If the publishers win, it could force AI companies to licence song catalogues before training future models, the same way a streaming service pays a fee every time a song is played. That would mean more money flowing back to composers and lyricists. A loss, or a settlement with modest damages, could set the opposite precedent.
For anyone who uses Claude or similar AI tools, nothing about the product changes while litigation runs its course. But the bill for training those systems, if courts decide there is one, could eventually affect how AI companies price their services.



