AI music startup Suno responds to record labels, admits to training on copyrighted music

Five weeks later After three major record labels filed lawsuits against AI-powered music startups Suno and Udio, accusing them of “trampling on the rights of copyright owners,” Suno responded in a legal filing Thursday, admitting that it trained its music-generation model on copyrighted music — while insisting that doing so was consistent with the fair use doctrine.

“What the major record labels really don’t want is competition,” Suno’s lawyers wrote in their response to the label’s complaint. “Where Suno sees musicians, teachers and everyday people using a new tool to create original music, the labels see a threat to their market share.” (Rolling Stone (referred to the RIAA, the industry association for film studios, for comment.)

More from Rolling Stone

Suno lets users create songs by simply typing in descriptions, and in its latest version, it also offers the option to extend its AI model to include uploaded audio files. According to Suno’s documentation, the company trained its neural network on “tens of millions” of recordings—all the music available on the “open web,” which would include millions of copyrighted songs available on YouTube and elsewhere. “It is no secret,” the documentation states, “that the tens of millions of recordings on which Suno’s model was trained likely included recordings owned by Plaintiffs in this case.” Previously, company executives and investors have suggested that copyrighted material was used in the training process, but have not directly confirmed it.

In the lawsuit, filed June 25, the labels didn’t claim that Suno’s music violates copyright laws—instead, they focused on the training process behind them. “Basic principles of copyright law dictate that copying protected sound recordings for the purpose of developing an AI product requires the consent of the rights holders,” the label’s lawyers wrote. “Suno cannot avoid liability for willful copyright infringement by relying on fair use. While the fair use doctrine promotes human expression by permitting unlicensed uses of copyrighted works in certain limited circumstances, Suno is offering machine-generated music that imitates human creativity and expression—not human creativity or expression.”

In response to the complaint, Suno’s lawyers call the studio’s argument “misleading” and say the focus on the training process “will prove fatal to Plaintiffs’ claims.” “Fair use under copyright law involves copying a protected work as part of a back-end technological process,” Suno’s response states, “invisibly to the public, for the purpose of ultimately creating a non-infringing new product. Congress enacted the first copyright law in this country in 1791. In the 233 years since, no case has ever—not once—concluded to the contrary.… The results Suno generates are new sounds, shaped precisely by the ‘styles, arrangements, and tones’ of those that preceded them. They are inherently lawful.”

Suno’s response also accuses the brands of “abusing copyright” to suppress competition and argues that they can win the case on that basis alone. “With the benefit of discovery, Suno will demonstrate that, even setting aside the fair use issue, Plaintiffs’ widespread copyright abuse completely precludes their claims,” the response states.

In a blog post accompanying the filing, Suno says its model learns from past music just as much as musicians do. It also calls the timing of the label lawsuit “somewhat surprising,” since the company had been in talks with major labels about “jointly expanding the music pie.” “We did this not because we had to,” the company writes in the blog post, “but because we believe the music industry can help us lead this expansion of opportunity for all, rather than resist it. Whether this lawsuit is the result of overzealous lawyers throwing their weight around or a deliberate strategy to gain leverage in our commercial discussions, we believe this lawsuit is an unnecessary obstacle to a larger and more valuable future for music.”

The debate over training data for generative AI models is broad and ongoing. New York Timesfor example, it sued OpenAI, the company behind ChatGPT, making almost identical claims to those in the big label lawsuit — including claiming that the process of copying Times“Uploading data to their servers during the training process was a copyright infringement.

The Best of Rolling Stone