Udio and Suno are not, despite their names, the most popular new restaurants on the Lower East Side. They're ai startups that allow people to generate stunningly real-sounding songs, complete with instrumentation and vocal performances, from prompts. And on Monday, a group of major record labels sued them, alleging copyright infringement “on an almost unimaginable scale,” alleging that the companies can only do this because they illegally ingested huge amounts of copyrighted music to train their artificial intelligence models.
These two lawsuits contribute to a growing number of legal headaches for the ai industry. Some of the most successful companies in the space have trained their models with data acquired through the ai-ignore-rule-scraping-web-contect-robotstxt”>unauthorized scraping of enormous amounts of information from the Internet. ChatGPT, for example, was initially trained in millions of documents collected from links posted on Reddit.
These lawsuits, led by the Recording Industry Association of America (RIAA), address music more than the written word. But as same as The New York Times' against OpenAI, they raise a question that could reshape the tech landscape as we know it: can ai companies just take what they want, turn it into a product worth billions, and claim it was fair use?
“That's the key problem that needs to be solved, because it affects all kinds of different industries,” he said. Paul Faklerpartner at the Mayer Brown law firm specializing in intellectual property cases.
What are Udio and Suno?
Both Udio and Suno are fairly new, but they've already made quite a splash. Suno was launched in December by a Cambridge-based team that previously worked for Kensho, another artificial intelligence company. It quickly partnered with Microsoft which integrated Suno with Copilot, Microsoft's ai chatbot.
Udio was launched just this year, ai-music-raises-10-million-65-million-for-spines-ai-more-sora-cinematic-ai/”>raising millions of dollars from heavyweights in the world of technology investing (Andreessen Horowitz) and the world of music (Will.i.am and Common, for example). The Udio platform was used by comedian King Willonius to generate “BBL Drizzy,” Drake's song that went viral after producer Metro Boomin remixed it and released it to the public so anyone could rap.
Why is the music industry suing Udio and Suno?
The RIAA's lawsuits use lofty language and say this litigation is about “ensuring that copyright continues to fuel human invention and imagination, as it has for centuries.” This sounds good, but ultimately the incentive he's talking about is money.
The RIAA states that generative ai poses a risk to the business model of record companies. “Instead of licensing copyrighted sound recordings, potential licensees interested in licensing such recordings for their own purposes could generate ai-like sound at virtually no cost,” the lawsuits state, and add that such services could “(flood) the market with 'imitators' and 'soundlikes,' thus revolutionizing an established sample licensing business.”
The RIAA is also seeking compensation of $150,000 per infringing work, which, given the huge bodies of data typically used to train ai systems, is a potentially astronomical figure.
Does it matter if ai-generated songs are similar to real ones?
The RIAA's lawsuits included examples of music generated with Suno and Udio and comparisons of their music notation to existing copyrighted works. In some cases, the generated songs had small phrases that were similar; for example, one began with the line sung “Jason Derulo” in the exact cadence with which Jason Derulo in real life begins many of his songs. Others had extended sequences of similar notation, as in the case of a song inspired by Green Day's “American Idiot.”
One track began with the line “Jason Derulo” sung in the exact cadence with which Jason Derulo in real life begins many of his songs.
This seems Pretty damning, but the RIAA does not claim that these specific soundtracks infringe copyright; rather, it claims that ai companies used copyrighted music as part of their training data.
Neither Suno nor Udio have made their training data sets public. And both companies are vague about the sources of their training data, although that's par for the course in the ai industry. (OpenAI, for example, has dodged questions whether YouTube videos were used to train their Sora video model).
The RIAA lawsuits note that Udio CEO David Ding has said the company trains with “better quality” music that is “publicly available” and that a Suno co-founder wrote on the official Suno Discord that the company trains with a “mix of proprietary and public data.”
Fakler said including examples and comparisons of notations in the lawsuit is “insane” and said it went “way beyond” what would be necessary to allege legitimate grounds for a lawsuit. For one, the labels may not own the composition rights to the songs allegedly ingested by Udio and Suno for training. Rather, they own the copyright to the sound recording, so showing similarity in musical notation doesn’t necessarily help in a copyright dispute. “I think it’s actually designed for optics for PR purposes,” Fakler said.
On top of that, Fakler noted, it's legal to create a similar audio recording if you own the rights to the underlying song.
When reached for comment, a Suno spokesperson shared a statement from CEO Mikey Shulman stating that their technology is “transformative” and that the company does not allow cues that name existing artists. Udio did not respond to a request for comment.
Is this fair use?
But even if Udio and Suno were to use the labels' copyrighted works to train their models, there's one big question that could override everything else: is this fair use?
Fair use is a legal defense that allows the use of copyrighted material in the creation of a significantly new or transformative work. The RIAA argues that the startups cannot claim fair use, saying that Udio and Suno's results are intended to replace real recordings, that they are generated for commercial purposes, that the copying was extensive rather than selective, and, finally, that The resulting product represents a direct threat to the label business.
In Fakler's view, startups have a strong fair use argument as long as copyrighted works are only copied temporarily and their defining characteristics are extracted and abstracted into the weights of an ai model.
“It's about extracting all that material, just like a musician would learn those things by playing music.”
“That’s how computers work: They have to make these copies and then the computer analyzes all this data to be able to extract the copyright-free material,” he said. “How do we construct songs that the listener can understand as music and that have various characteristics that we commonly find in popular music? It’s about extracting all those things, just as a musician would learn those things by playing music.”
“In my opinion, that's a very strong fair use argument,” Fakler said.
Of course, a judge or jury may not agree. And what is brought to light in the discovery process (if these lawsuits get there) could have a big effect on the case. What music tracks were taken and how they ended up in the training set could matter, and specific details about the training process could undermine a fair use defense.
We're in for a long ride as the RIAA lawsuits and others like them make their way through the courts. From text and photographs to sound recordings, the question of fair use looms over all of these cases and the ai industry as a whole.
