On Monday, June 24, 2024, the three U.S. major record labels, Universal Music Group, Warner Music Group and Sony Music Entertainment, filed copyright infringement claims against two artificial-intelligence (AI) music start-ups, Suno and Udio. These AI music companies allow users to generate musical recordings based on text prompts, including by genre and topic. The two lawsuits add to the growing list of major litigation arising out of the use of proprietary material to train generative AI models, but represent collective action by the major industry players in the music industry in a way that other creative industries (such as film and television) have thus far largely eschewed.
The basis for the claims in both lawsuits is that Suno and Udio each used “reams of [c]opyrighted [r]ecordings” either owned or exclusively controlled by the labels to train their AI models, resulting in unauthorized copying in multiple instances throughout the process. The complaints do not allege claims based on a right of publicity theory (though they do note the presence of files generated with recognizable vocals), nor do they focus on the underlying musical compositions embodied in the in the copyrighted recordings. Rather, the claims appear designed to straightforwardly target the use of copyrighted sound recordings alone.
The labels provided several illustrative examples of instances in which the new recordings generated in response to text prompts on both platforms contained components that were exactly the same as the copyrighted materials, supporting their argument that the platforms “could not have built a model capable of producing audio so similar to the [c]opyrighted [r]ecordings without the initial act of copying those recordings.” The examples cited include the generation of outputs that mimic the style and content of Chuck Berry’s “Johnny B. Goode,” Bill Haley & His Comets’ “Rock Around the Clock,” B.B. King’s “The Thrill Is Gone,” Green Day’s “American Idiot,” Mariah Carey’s “All I Want for Christmas Is You,” and others. The suit against Suno also cited the presence of recognizable producer tags in outputs, which allegedly show the training from ingested copyrighted recordings.
In addition to providing support for their claim of unauthorized copying, the labels anticipated and offered prebuttals of the defendants’ potential fair use defenses, arguing that such defense will fail because their use of copyrighted sound recordings is purely for commercial gain and would use human-created works to supplant human creation. In response, Suno’s CEO issued a statement that adopted the terminology of a fair use defense, stating, “Our technology is transformative; it is designed to generate completely new outputs, not to memorize and regurgitate pre-existing content.”
The labels seek (1) declarations that Suno and Udio used copyrighted sound recordings to train their AI models without permission; (2) injunctions to prevent Suno and Udio from using copyrighted sound recordings to train their AI models; and (3) statutory damages in the amount of $150,000 per sound recording infringed. Unlike in the major music publishers’ lawsuit against Anthropic, the complaints do not reference a request preliminary injunctive relief.
The lawsuits also appear to be a reaction to start-ups’ habit of asking for forgiveness not permission in building new technology that incorporates or is based on proprietary works, especially musical works. This has often resulted in after-the-fact settlements and go-forward licenses, as were the cases with Peloton and TikTok. Unless the labels seek preliminary injunctive relief here, their strategy may be to force the start-ups into compliance and licensing and to send a message to other would-be AI-training infringers. It is also possible that fair use determinations in earlier-filed litigation will inform any resolution over the fair use defense in these cases, should a settlement not be reached first.