Major US Recording Labels Sue AI Music Start-ups

On Monday, the Recording Industry Association of America (RIAA) announced a lawsuit from major recording labels alleging copyright infringement from artificial intelligence music sites. Sony Music Entertainment, Universal Music Group, and Warner Music Group have teamed up to take down start-ups Suno and Udio. Both have essentially the same function: generating a song from a user's text prompt such as Suno's suggested “heartfelt Samba song about a photo on the mantel”. Both also display trending user-generated music across a wide range of genres, themes, and even languages.

This novel and fascinating technology doesn't spring from nothing, though. The models for Suno and Udio have to have been trained with large amounts of data to recognize the patterns of music and song that distinguish what the models produce from random spurts of word and sound. The labels claim that Suno and Udio copied “decades worth of the world's most popular sound recordings” to train their AI models on, partaking in “willful copyright infringement at an almost unimaginable scale”.¹ The lawsuit aims to have Suno and Udio admit to their infringement of the labels' copyrighted recordings, be barred from using their recordings for training in the future, and pay damages of up to $150,000 per song already used.

What responses have Suno and Udio given?

In correspondence, Suno and Udio both claimed “fair use” of the copyrighted materials.² While determining fair use is very case dependent, the U.S. Copyright Office writes that courts are “more likely to find that nonprofit educational and noncommercial uses are fair,” as compared to commercial uses.³ While Suno and Udio are certainly for-profit, their use may also qualify as fair by being ‘transformative', as defined by the U.S. Copyright Office as “with a further purpose of different character” as opposed to a copy or imitation of the original copyrighted work. A blog post on Udio's website further explains the start-up's thinking: “Our system is explicitly designed to create music reflecting new musical ideas. We are completely uninterested in reproducing content in our training set, and in fact, have implemented and continue to refine state-of-the-art filters to ensure our model does not reproduce copyrighted works or artists' voices”.⁴ However, whether Udio's statements hold true is in question, as lawyers representing the recording labels claim that both websites could imitate copyrighted recordings, and even specific artists like ABBA and Jason Derulo.

Suno's CEO, Mikey Shulman, released a statement in response to the lawsuit, saying that Suno is “designed to generate completely new outputs, not to memorize and regurgitate pre-existing content,” and also noting that users are not allowed to reference specific artists.⁵ If Suno and Udio are able to prove that their models are truly ‘transformative' their fair use claim may stand and set the precedent that using copyrighted content to train AI models is not a form of infringement. If that is the ruling, AI music's burgeoning will likely continue and accelerate, but if not, regulation and restrictions may put a damper on that growth.

Both start-ups seem to avoid some essential information, though. The RIAA responded to Shulman's statement with the simple question of “what sound recordings have they illegally copied?”⁶ Neither Suno nor Udio has publicly released the data their models were trained on, and that lack of transparency leaves suspicion and distrust.

Ethical concerns

On Udio's blog post, the AI model is metaphorized as a music student, listening to and studying music to learn a multitude of techniques.⁷ What this comparison misses is that students generally have to pay for others' expertise, such as through music teachers, music books, and sheet music, while their AI model has used the work of multitudes of artists without any compensation or permission. RIAA emphasized this disparity, saying that Suno and Udio's claiming “it's ‘fair' to copy an artist's life's work and exploit it for their own profit without consent or pay [sets] back the promise of genuinely innovative AI for us all.”⁸

This lawsuit isn't the first case of artists decrying the devaluation of their work and denial of fair compensation by AI companies. In April, over 200 artists signed an open letter urging those in AI to not infringe upon the rights of human artists.⁹The battle between protecting artists' work and widening music accessibility through technology will likely continue. A fundamental question underlies the debate; is what Suno and Udio produce really music and are the individuals who input prompts artists? Many don't think so, but those who do see Suno and Udio as creating new and widespread opportunities for a new art form. From that perspective, the resistance to AI-generated music can be viewed like the resistance to photography when it was invented. Upon seeing a photograph for the first time in 1840, Paul Delaroche, a French painter, declared that “from today, painting is dead”.¹⁰ We know now that the birth of photography did not mark the death of painting, and that since then, painting has continued to develop and evolve as simply a separate medium. In that same sense, AI-generated music may be able to coexist with human-made music in a way that allows both to flourish. Suno and Udio may not yet have the right balance for that, but that doesn't mean it can't happen.