Sony Music Publishing and Warner Chappell Music are among a group of music publishers suing artificial intelligence company Anthropic, escalating the music industry’s legal battle over how copyrighted material is used to train generative AI systems.
The complaint was filed on 28 August 2026 in the US District Court for the Northern District of California. According to Reuters’ reporting on the lawsuit, the publishers accuse Anthropic, the company behind the Claude AI models, of improperly obtaining and using copyrighted musical compositions during AI training.
What the music publishers are alleging
According to the complaint, copyrighted song lyrics and sheet music were allegedly obtained through sources including torrent downloads and then used in the development of Claude.
The publishers allege that the material includes works associated with major artists and songwriters including The Beatles, Taylor Swift and Michael Jackson. They also claim Claude is capable of reproducing portions of copyrighted lyrics when prompted.
These remain allegations before the court. No finding has been made that Anthropic infringed the publishers’ copyrights.
Anthropic says it will defend the case
Anthropic has rejected the allegations and said it intends to defend itself. The company maintains that training artificial intelligence systems on copyrighted material can qualify as fair use under US copyright law.
The publishers are seeking statutory damages that could reach up to US$150,000 for each copyright found to have been infringed, as well as a court order preventing Anthropic from using the affected works.
Why this case matters to the music industry
The dispute goes beyond whether an AI chatbot can reproduce song lyrics. One of the central questions is how copyrighted works were acquired in the first place and whether AI companies can lawfully use those works as training material without obtaining licences from their owners.
That distinction could become increasingly important for musicians, composers and publishers. The industry is simultaneously dealing with AI-generated recordings, synthetic artists and automated songwriting. At the same time, listeners themselves are finding it increasingly difficult to identify synthetic recordings by ear. A recent blind test found that 97% of listeners could not reliably distinguish fully AI-generated music from human-made music.
Copyright is only one part of the wider debate. Transparency is another. Music brands are also being questioned about how generative tools are used in commercial content, as demonstrated when D’Addario acknowledged that generative AI had been used in a controversial product demonstration.
A legal battle that could shape AI music
The Anthropic lawsuit joins a growing group of copyright cases attempting to establish where the legal boundaries of generative AI training should sit.
The outcome could influence how future AI systems license, document and potentially compensate creators for the music and other copyrighted material used during their development.
For musicians and songwriters, the larger question is becoming increasingly difficult to ignore: as generative AI becomes more capable, who should have the right to decide when copyrighted creative work can be used to help build it?
