Generative AI platforms Udio and Suno have taken aim at the arguments introduced by the majors in a pair of infringement suits. Photo Credit: BoliviaInteligente
AI-powered music-creation platforms Suno and Udio have officially fired back against the high-stakes copyright infringement lawsuits they’re facing from the major labels.
Both defendants just recently took aim at the suits; Udio is being sued in a New York federal court, while Suno is facing a separate-but-similar action in Massachusetts. We’ve covered the cases – and a public war of words between the defendants and the RIAA – in detail.
At the top level, though, they revolve around the all-important question of whether companies have the legal authority to copy and train generative AIs on protected materials (which are then incorporated into outputs) without the authorization of rightsholders.
Predictably, Suno and Udio are adamant that they do possess the authority, with the training process at hand purportedly drawing from the basic “building blocks of music” and constituting fair use, per their filings.
Putting everything out in the open, the AI companies in their answers directly acknowledged that both models utilized the majors’ recordings to train.
“The many recordings that Udio’s model was trained on presumably included recordings whose rights are owned by the Plaintiffs in this case,” reads one relevant line.
But doing so is lawful under copyright law, the responses claim in more words, including not only when recordings are copied behind the scenes but, more than that, when the AI outputs share characteristics with protected works.
“Under longstanding precedent,” Udio and its counsel wrote on the former front, “it is fair use to make a copy of a protected work as part of a back-end technological process, invisible to the public, in the service of creating an ultimately non-infringing new product.”
And on the equally important output side, the defendants underscored the belief that copyright law, owing to a described carve-out of sorts for recordings resembling existing works but not copying directly, permits their soundalike outputs.
“Even to the extent that Udio’s outputs ‘imitate or simulate’ sounds in the Plaintiffs’ recordings,” Udio penned, “Congress made the public policy choice to immunize such new creations from copyright infringement liability, so long as they do not themselves contain actual snippets of pre-existing recordings. Which they do not.”
Taking that important argument a step further – and using as an example Frank Sinatra’s “My Way,” which the majors say has been infringed upon because an output allegedly shared melodic characteristics with the famed recording – Udio emphasized the “literally hundreds of different recordings” of the work available on streaming platforms.
“Consequently,” Udio relayed, “Plaintiffs’ argument betrays a profound misunderstanding of the technology at issue by suggesting that UMG’s particular version must have been in the training set because Udio allegedly generated an output that contains ‘melodic similarities to the Sinatra original throughout.’ So too do countless other recordings of the song.”
Attempting to replicate the recording with a prompt containing the “My Way” lyrics, Udio added for good measure, allegedly violated its terms of use.
“No one owns musical styles,” the platform concluded. “Developing a tool to empower many more people to create music, by analyzing on a massive scale the relationships among notes and rhythms and tones to ascertain the building blocks of different musical styles, is a quintessential fair use under longstanding and unbroken copyright doctrine.”
Suno responded to the RIAA-spearheaded suit with similar arguments, and both also took the opportunity to point out the majors’ perceived “aversion to competition,” depicting generative AI as the latest in a line of resisted but ultimately accepted innovations.
Furthermore, “the major record labels wield massive market power” and haven’t “hesitated to exploit it in fundamentally anticompetitive ways,” according to both answers, which maintain that the alleged anticompetitive behavior is carrying over to dealings within the AI sector.
The RIAA reached out with a roughly 250-word statement about the responses, seizing on the above-described training admissions and doubling down on its position.
“After months of evading and misleading,” an RIAA spokesperson communicated in part, “defendants have finally admitted their massive unlicensed copying of artists’ recordings. It’s a major concession of facts they spent months trying to hide and acknowledged only when forced by a lawsuit.
“Their industrial scale infringement does not qualify as ‘fair use’. There’s nothing fair about stealing an artist’s life’s work, extracting its core value, and repackaging it to compete directly with the originals, as the Supreme Court just held in its landmark Warhol Foundation case.
“Defendants had a ready lawful path to bring their products and tools to the market – obtain consent before using their work, as many of their competitors already have. That unfair competition is directly at issue in these cases.”
