Music publishers Universal Music Publishing, Concord and ABKCO have sued AI company Anthropic for copyright infringement through the US courts, accusing the tech firm of copying and exploiting lyrics that they control without licence.
The publishers reckon that, while Anthropic's technology may be "complex and cutting edge", the copyright arguments in their lawsuit are super simple. Indeed, they say, a cursory glance of the Statue Of Anne of 1710 - which kickstarted English copyright and went on to influence other Anglo-American copyright regimes - will demonstrate how Anthropic is liable for copyright infringement.
“A defendant cannot reproduce, distribute and display someone else’s copyrighted works … unless it secures permission from the rightsholder”, they state. “This foundational rule of copyright law dates all the way back to the Statute Of Anne in 1710, and it has been applied time and time again to numerous infringing technological developments in the centuries since”.
"Anthropic builds its AI models by scraping and ingesting massive amounts of text from the internet and potentially other sources", it goes on, using "that vast corpus to train its AI models and generate output based on this copied text".
Crucially, "included in the text that Anthropic copies to fuel its AI models are the lyrics to innumerable musical compositions for which publishers own or control the copyrights".
"As a result of Anthropic’s mass copying and ingestion of publishers’ song lyrics", the lawsuit goes on, "Anthropic’s AI models generate identical or nearly identical copies of those lyrics, in clear violation of publishers’ copyrights".
It then claims that if a user asks Anthropic’s Claude AI chatbot to provide lyrics to a song owned by one of the publishers, "the chatbot will provide responses that contain all or significant portions of those lyrics". And while Google search and lyric websites do the same, they - of course - are licensed by the music industry.
Not only that, "Anthropic’s AI models generate output containing publishers’ lyrics even when the models are not specifically asked to do so". So a user might ask the AI to write a song about a certain topic or produce some fiction in the style of a certain artist, and Claude responds by "generating output that nevertheless copies publishers’ lyrics".
For copyright owners, like music publishers, it is clear that if a tech firm trains an AI with existing copyright protected content they must get permission from the relevant copyright owners. However, some AI companies reckon such training is covered by copyright exceptions or, in the US, the principle of fair use.
There are already a number of cases working their way through the courts that will test the copyright obligations of AI companies, though to date those have mainly been filed by the owners of visual or literary copyrights, so it's interesting to see a music case join the party.
Commenting on the new lawsuit, Matthew J Oppenheim, one of the lawyers representing the publishers, says: “The unauthorised use of copyrighted material is illegal and, in the case of copyrighted music lyrics, harms songwriters and music publishers”.
“It is well established by copyright law that an entity cannot reproduce, distribute and display someone else’s copyrighted works to build its own business unless it secures permission from rightsholders”, he adds. “Just like countless other technologies, AI companies must abide by the law".