Sony Music Publishing and Warner Chappell Music have sued artificial intelligence company Anthropic, accusing it of illegally using copyrighted music and other intellectual property to develop and train its Claude AI models. The publishers have described the alleged conduct as one of the “largest and most blatant” cases of intellectual property theft.
The lawsuit, filed in the US District Court for the Northern District of California, names Anthropic as well as its co-founders Dario Amodei and Benjamin Mann as defendants. The music publishers allege that Anthropic engaged in large-scale copying, scraping and downloading of copyrighted works without obtaining permission from rights holders.
At the centre of the dispute is the use of copyrighted musical works, particularly song lyrics, in the development of Anthropic’s AI systems. According to the complaint, Anthropic allegedly used tens of thousands of musical works without authorization. The publishers argue that such use not only violates copyright protections but also allows the AI company to commercially benefit from works created by songwriters and artists.
The lawsuit alleges that Anthropic obtained copyrighted material through multiple sources, including unauthorised online repositories and torrent-based services. The publishers claim that the company and its employees downloaded millions of copyrighted books and other materials, which were subsequently used as part of the data infrastructure supporting its AI models.
The complaint also alleges that Anthropic scraped copyrighted song lyrics from licensed websites and other online sources. The publishers contend that these lyrics were incorporated into the training process for Claude, enabling the system to reproduce or closely imitate protected material when prompted by users.
Among the works allegedly involved are well-known songs such as “Livin’ on a Prayer,” “Ain’t No Mountain High Enough” and “Hallelujah.” The publishers argue that the issue extends far beyond individual songs, pointing to what they describe as the systematic use of large quantities of copyrighted material.
The case represents another major confrontation between the creative industries and the rapidly expanding artificial intelligence sector. AI companies depend on enormous datasets to train large language models, but the practice has raised questions about whether copyrighted material can be used without the permission of creators and rights holders.
Music publishers have argued that the scale of AI training makes the issue particularly significant for the music industry. Songwriters and publishers spend years creating, producing and licensing musical works, while AI systems can potentially absorb large collections of those works and use them to generate content or respond to users.
The publishers are seeking statutory damages of up to $150,000 for each infringed musical work. They are also seeking additional damages in connection with allegations that copyright management information was removed from protected works. Depending on the number of works ultimately found to have been infringed, the financial consequences for Anthropic could run into billions of dollars.
The latest lawsuit comes as Anthropic is already facing substantial legal scrutiny over its use of copyrighted material. The company has previously been sued by authors and music publishers over allegations that copyrighted works were used to train its AI systems without permission.
In an earlier copyright dispute involving authors, Anthropic agreed to a $1.5 billion settlement. The case also raised an important distinction between the use of copyrighted works for AI training and the manner in which those works were obtained. While the use of copyrighted material for training has been the subject of continuing debate over fair use, the acquisition of material through piracy has faced significantly greater legal challenges.
The music publishers’ latest action therefore places renewed focus on how AI companies acquire training data. The dispute is not limited to whether copyrighted material can be processed by an AI model. It also raises questions about whether companies can lawfully obtain such material from unauthorised sources and whether AI-generated responses that reproduce protected content constitute infringement.
Anthropic has disputed the allegations and is expected to defend itself in court. The company has previously introduced measures designed to prevent Claude from generating copyrighted song lyrics in response to user requests. However, the publishers argue that safeguards introduced at the output stage do not resolve the underlying issue of how copyrighted works were acquired and used during the development of the models.
The lawsuit could have implications well beyond Anthropic and the music industry. Several major technology companies are currently facing copyright disputes involving books, news articles, music, images and other forms of creative content used to develop generative AI systems.
For rights holders, the lawsuits represent an effort to establish clearer boundaries around the use of creative works in AI development. For AI companies, the cases could determine how training datasets can be assembled, what constitutes lawful use of copyrighted material and what obligations companies have towards creators.
The Sony and Warner case is therefore likely to become another closely watched legal battle in the evolving relationship between artificial intelligence and intellectual property. As AI systems become increasingly capable of generating and reproducing content, courts will face growing pressure to determine where technological innovation ends and copyright infringement begins.
The outcome could influence future licensing models, the cost of AI training data and the way technology companies work with artists, writers, publishers and other creators. More broadly, the case highlights a central question facing the AI industry: how can companies build increasingly powerful systems while respecting the rights of the people whose creative work helps make those systems possible?
