Universal Music Publishing Group and two other music publishers are suing Anthropic, the company behind the artificial intelligence chat bot Claude, for copyright infringement.
In a Complaint filed with in the United States District Court for the Middle District of Tennessee in Nashville on October 18, 2023, the publishers claim that Claude was trained on vast amounts of unlicensed song lyrics, which Claude can output either in their original form or as the foundation for new works, all without any payment to the lawful rightsholders.
The publishers point out that when prompted Claude can output word-for-word copies of the lyrics of Katy Perry’s “Roar,” Gloria Gaynor’s “I Will Survive,” and Garth Brooks’ “Friends in Low Places.” In addition, they note that Claude can create new works drawing heavily from Don McLean’s “American Pie,” the theme song to “The Fresh Prince of Bel-Air,” Louis Armstrong’s “What a Wonderful World,” Steppenwolf’s “Born to Be Wild,” and Johnny Cash’s “Daddy Sang Bass” (including chords).
There are several issues at play. One issue is whether it is fair use to train a large language model or other AI on copyrighted works without proper rights and clearances. Another issue is to what extent the outputs of a large language model (LLM) are “derivative works” that may infringe original works under copyright law. Yet another issue is whether rightsholders must be compensated for their works being ingested as training materials for an LLM (and, if so, how much). There are many more issues beyond these. This lawsuit along with others filed against OpenAI, Meta, and Stability AI (to name a few) hope to answer these questions.
In anticipation of plenty more litigation to come, tech giants Google, Microsoft, and Adobe have all committed to indemnify users for copyright infringement claims that copyright owners bring users regarding the outputs of their generative AI models. The tech giants’ indemnification policies allow them to assert the public position that they stand behind their technology. At the same time, these indemnification policies enable them to strategically ensure that their choice of litigation counsel argues their positions on the issues that will shape their industry for decades to come.
Whether you are building an AI or using an AI, you should stay tuned to these issues. I look forward to reporting on these issues as they develop.
Comments