Universal Music Group filed a copyright infringement lawsuit against artificial intelligence startup Anthropic on Wednesday citing widespread scraping of its clients’ song lyrics being used to train its ChatGPT rival chatbot Claude.

This marks the first major lawsuit by the music industry against an AI developer and UMG is seeking $75 million in damages. Other publishers joining UMG in the lawsuit included Concord Music Group, ABKO, Worship Together Music and CMG, among others.

“As a result of Anthropic’s mass copying and ingestion of publishers’ song lyrics, Anthropic’s AI models generate identical or nearly identical copies of those lyrics, in clear violation of publishers’ copyrights,” the complaint states.

It goes on to note that the AI models will generate output containing lyrics when not prompted to and even write songs about certain topics and provide chord progressions and musical compositions and fiction that match those of certain artists and songwriters.

As an example, the Claude chatbot can be asked to repeat the lyrics of “Every Breath You Take,” written by the group the Police. UMG bought the rights to the song, along with all of Sting’s songwriting catalog in 2022, as part of a deal estimated to be worth $300 million.

Although there are websites on the internet that aggregate and provide song lyrics on demand, they are licensed to do so by the publishers. The lawsuit says that Anthropic did not seek to license itself to use the song lyrics as part of training its model and the fact that the model produces the lyrics means that it has unlawfully ingested them.

“By refusing to license the content it is copying and distributing, Anthropic is depriving publishers and their songwriters of control over their copyrighted works and the hard-earned benefits of their creative endeavors,” the lawsuit says.

This lawsuit follows questions about how AI large language models are trained and what kind of data and sources they should incorporate. Many AI developers consume large swaths of the internet in order to train models to help them produce more lifelike conversational messages and to keep up with popular knowledge.

Anthropic is the latest AI developer to be targeted with a lawsuit involving copyright infringement. ChatGPT developer OpenAI LP and Mark Zuckerberg’s Meta Platforms Inc. also faced a lawsuit from comedian and author Sarah Silverman and two other authors in July claiming that the companies used copyrighted materials from their books when training their AI chatbots.

Similar to the UMG lawsuit, the complaint cited that the chatbots could summarize the books written by the plaintiffs, showing that the models had been trained on data taken from their pages.

The same lawyers who filed that lawsuit, sued the art-generating AI providers Stability AI Ltd., Midjourney Inc. and DeviantArt Inc. on behalf of three artists alleging that their artwork had been used without their permission. Separately, Getty Images also sought a lawsuit against Stability AI, claiming that it used more than 12 million copyrighted images in its training set.

Image: Pixabay