Record labels sue two startups for training AI models on their songs

The world’s biggest record labels are suing two artificial intelligence startups, taking an aggressive stance to protect their intellectual property against technology that makes it easy for people to generate music based on existing songs.

The Recording Industry Association of America said it filed twin lawsuits Monday against Suno AI and Uncharted Labs Inc, the developer of Udio AI, on behalf of Universal Music Group NV, Warner Music Group Corp and Sony Music Entertainment. The complaints allege the companies are unlawfully training their AI models on massive amounts of copyrighted sound recordings.

The RIAA, a trade group for record labels, is seeking damages of as much as US$150,000 “per work infringed”. That could amount to potentially billions of dollars.

“The music community has embraced AI, and we are already partnering and collaborating with responsible developers to build sustainable AI tools centered on human creativity that put artists and songwriters in charge,” Mitch Glazier, chief executive officer of the RIAA, said in a statement.

“But we can only succeed if developers are willing to work together with us. Unlicensed services like Suno and Udio that claim it’s ‘fair’ to copy an artist’s life’s work and exploit it for their own profit without consent or pay set back the promise of genuinely innovative AI for us all.”

Suno and Udio are among a new crop of startups that use generative AI to automate the music-making process. People can type in a short written prompt, like “an electro-pop song about strawberries,” and software from either company will spit out human-sounding music in seconds. In order to build their AI systems, the companies must first train their software on enormous datasets, which can be made up of many millions of individual pieces of information.

Suno’s technology is “transformative”, and is “designed to generate completely new outputs, not to memorise and regurgitate pre-existing content”, co-founder and CEO Mikey Shulman said in a statement. That’s why the company doesn’t allow users to include the names of musical artists in their written prompts when creating songs, he explained.

“We would have been happy to explain this to the corporate record labels that filed this lawsuit (and in fact, we tried to do so), but instead of entertaining a good faith discussion, they’ve reverted to their old lawyer-led playbook,” he said.

Udio didn’t respond to a request for comment.

Founded in 2022, Cambridge, Massachusetts-based Suno first released its music-making software last year and raised US$125mil in May. Udio, started by former Google DeepMind researchers and engineers and based in New York, introduced a “beta” version of its software in April and raised US$10mil in funding. Both allow users to make some songs for free while also offering monthly subscriptions to those who want to create more.

The legal challenge from the music industry is just the latest example of technology colliding with creative industries as generative AI is increasingly used to churn out all kinds of content. Companies like Midjourney, OpenAI and Stability AI built their media-generating AI models with datasets that pull imagery from across the Internet. While they argue that the practice is protected under the fair use doctrine of US copyright law, it has led to outrage and lawsuits .

In the music industry, artists and labels see AI as a potential existential threat. Hundreds of musicians, including Billie Eilish, Miranda Lambert and Aerosmith signed an open letter in April via the nonprofit Artist Rights Alliance, urging AI developers, tech companies and others to halt the use of AI “to infringe upon and devalue the rights of human artists.” At the same time, record labels are scrambling to balance the creative potential of the fast-moving technology while also protecting artists’ rights and their own profits

“There is both promise and peril with AI,” according to the complaint against Udio. “As more powerful and sophisticated AI tools emerge, the ability for AI to weave itself into the processes of music creation, production, and distribution grows. If developed with the permission and participation of copyright owners, generative AI tools will be able to assist humans in creating and producing new and innovative music.”

But those same tools, if not deployed responsibly, also risk creating “irreparable harm” to artists, labels and the industry, “inevitably reducing the quality of new music available to consumers and diminishing our shared culture.”

Neither Suno nor Udio would say what, precisely, their AI systems are trained on when asked by Bloomberg News in April. Udio co-founder and CEO David Ding said the company used publicly available data from the internet.

“We try to cast our net as wide as possible so that we can represent all the different traditions of music in our model,” Ding said at the time.

Shulman said in April that the training data is, in some ways, even more important than how the company constructs its AI software, “so we’re pretty closely guarding that secret.” Shulman also said Suno’s practices are “legal” and “fairly in line with what other people are doing.”

Remarkably similar

Music generated with the services can, at times, sound remarkably similar to copyrighted music. Ed Newton-Rex, CEO of nonprofit organisation Fairly Trained, which provides certification for AI models trained on licensed data, said he found it easy to generate a slew of tunes using both companies’ software that sound very much like artists such as Queen, Abba, Oasis, Blink-182 and Ed Sheeran.

In Monday’s lawsuits, the RIAA claims authentic producer tags appear on some of the music coming out of Suno and Udio, and that people who use the services have generated sounds very similar to numerous artist-made songs, including The Temptations’ My Girl, Green Day’s American Idiot and Mariah Carey’s All I Want For Christmas. They have also produced vocals that are indistinguishable from famous recording artists, including Lin-Manuel Miranda, Bruce Springsteen and Michael Jackson, according to the RIAA.

Shulman said Suno was thinking about how to eventually compensate musical artists for their work, but that “right now there’s just no good way to do that.”

“We work very closely with lawyers to make sure that what we’re doing is legal and industry standard,” he said in April. “If the law changes, obviously we would change our business one way or the other.”

The music industry aims to get out in front of the technology before it’s too late and has already put AI startups on notice.

UMG’s music publishing arm sued Anthropic, a generative AI company, in October over similar claims, focused in particular on the alleged copying of lyrics. In May, Sony Music sent a letter to more than 700 AI companies and streaming services warning them not to use the label’s copyrighted material without explicit permission and licensing. It said it had reason to believe its content had already been used to train, develop or commercialise AI systems without its permission.

The RIAA alleges that Suno and Udio, whether through an investor or in-house executives, have essentially admitted to using copyrighted material to develop their models.

An early investor in Suno said he probably wouldn’t have invested in the startup if it had deals with labels when it started, according to the complaint. He said that defending against lawsuits was a necessary risk.

Generative AI companies have plausible fair use defences for using works as training data, said Pamela Samuelson, a digital copyright expert and law professor at the University of California, Berkeley. But she said courts might look at music differently than they would other works such as computer code, text or images.

“The data type might actually matter,” Samuelson said. “I could see courts distinguishing based on that.”

The case against Suno was filed in the US District Court for the District of Massachusetts and the case against Uncharted Labs was filed in the US District Court for the Southern District of New York. – Bloomberg