(Image courtesy of Getty Images. “Suno, an AI music generator, is in legal hot water.”)
Ronak Lamba
Connector Contributor
Sony Music Entertainment and Universal Music Group have sued Suno following the launch of Suno v6, their newest AI music model. This is not the first time these two companies have sued Suno. In 2024, Sony and UMG were joined by Warner Music Group in a lawsuit against the AI music generation platform, on the basis that the methods in which Suno uses to train their AI models infringes on copyrighted sound recordings. While the lawsuit is still active, Warner Music Group is the only company among the three to have since partnered with Suno, according to Music Business Worldwide.
The new lawsuit claims that the new v6 model still violates copyrighted works belonging to the record companies. In the complaint, also provided by Music Business Worldwide, the plaintiffs asserted that for Suno “to produce convincing imitations of genuine sound recordings,” it starts with “copying a vast range of sound recordings,” which includes an “enormous volume” of copyrighted songs. They claim that over 61,000 copyrighted recordings were unlawfully used by Suno. The record companies also alleged that Suno gathered a large portion of these sound recordings through the unlawful act of “stream ripping” from Youtube. They further explained, “Suno did not delete the sound recordings it copied to train its models once training was complete. Instead, Suno retained—and continues to retain—copies of those recordings, including Plaintiffs’ Copyrighted Recordings, in one or more stored datasets.”
Suno hasn’t backed down. While they admit to obtaining data from Youtube for training purposes, they maintain that the training of their models adheres to the “fair use” doctrine, according to a report by Variety. Suno further stated that previous models that were trained on copyrighted data are no longer supported, asserting that their new model is trained on “creations” from their own users and community. However, these “creations” were made with the previous models, which leads Sony and UMG to believe that the new v6 model continues to infringe on their copyrighted works.
Major record labels aren’t the only ones taking action against Suno. SOCAN, the leading performing rights organization in Canada (such as ASCAP or BMI in the US), filed a lawsuit against Suno in early September, claiming the platform “[reproduces] copyrighted works represented by the Canadian music rights organization without permission or compensation,” according to Rockum. SOCAN supported their claim with specific examples. In one instance, they claimed that “Both Sides Now,” a hit by the legendary Canadian singer-songwriter Joni Mitchell, was the clear source material behind a Suno-generated output that was ‘‘virtually identical” in melody, harmony and lyrics.
Many artists are also enraged. In the same week of the SOCAN lawsuit, Jason Isbell, an independent Americana singer-songwriter, was joined by three other musicians in filing a class action lawsuit against Suno. While the previous lawsuits revolved around copyright infringement, Isbell’s side claims Suno enables an “unlawful imitation of the personal styles and voices of specific artists,” according to the New York Times. Mikey Shulman, CEO and co-founder of Suno, says, “we intentionally chose not to use artist names in our training metadata because our goal is to help people create original songs — not music that sounds like existing artists.” While Suno denies the ability for users to create prompts using names of actual artists, it was stated in the lawsuit that users can circumvent this by “simply [typing] in those artists’ names with spaces separating each letter (‘t a y l o r s w i f t’).”
In the complaint, the plaintiffs describe that Suno’s models reflect a certain “phenomena” found in machine learning: overfitting. “An AI model is ‘overfitted’ when it is too closely adapted to the data on which it was trained, making it difficult for the model to generalize to new data sets.” They continued, “if a user inputs the prompt ‘a jazz song about New York’ into an overfitted AI model, the model may output a file that closely resembles one of the jazz tracks on which it trained. Suno designed its product in this way, as it could not create a viable product if it was incapable of generating outputs that users find appealing.
