Click here to read the full article.
German music collection society GEMA notched another major legal victory in its campaign to rein in AI companies’ use of copyrighted material on Friday, winning the case it brought against the generative music platform
Suno.
The Munich Regional Court has ruled that Suno is in breach of German and U.S. copyright law by training its AI systems on songs from GEMA’s repertoire and storing and reproducing them.
While specifics of the ruling from the Munich Regional Court German legal proceedings were not made public, its decision requires AI companies to pay to license material “for the systematic use of GEMA’s repertoire and its commercial exploitation,” a press release from GEMA reads. The decision covers that use across both model training and output generation.
“Today, the Chamber made one thing crystal clear: AI models built on stolen intellectual property have no protection under the law. AI service providers must pay for licenses rather than helping themselves to our members’ works free of charge,” a statement from GEMA’s CEO, Dr Tobias Holzmüller, reads. “Today, the court’s verdict has significantly strengthened Europe’s position as a cultural centre.”
The society had accused the Massachusetts-based Suno of infringement over just six songs – “Atemlos,” “Daddy Cool,” “Rasputin,” “Big in Japan,” “Forever Young,” and “Mambo No. 5.” – which GEMA said it had been able to essentially reproduce on Suno through basic prompting, despite the platform’s purpose and business model being to generate “new” sound-based material for users.
GEMA, which has around 100,000 members, also argued that Suno “ingested” those members’ works to train its models, but did not license the material – which Suno had partially admitted to back in 2024.
“Suno’s training data includes essentially all music files of reasonable quality that are accessible on the open internet, abiding by paywalls, password protections, and the like, combined with similarly available text descriptions,” Suno wrote in a 2024 court filing in a different case, brought against it in the U.S. by the three major labels, Universal Music Group, Sony Music and Warner Music Group. (The latter reached a licensing agreement with Suno last Nov., but the other two majors remain plaintiffs in the case, which should be decided sometime next year.
Suno did not immediately respond to a request for comment.
The victory comes roughly nine months after GEMA won another important AI-related lawsuit against OpenAI, which was found to have trained its models on unlicensed lyrics. GEMA was awarded unspecified damages in the OpenAI case, but the organization’s overarching goal has been to establish a legal framework in the E.U. that reckons with AI companies’ prior use of copyrighted material and creates a path for that usage moving forward. To that end, GEMA has released a proposed licensing system for AI companies’ use of music, as well as a charter laying out several approaches to peoples’ use of AI generally.
In large part, AI companies’ defense has been that their ingestion of all this material – books, videos, music, articles, visual art and more – in order to train the models that power their platforms has all fallen under the umbrella of fair use. In the U.S., fair use allows for the use of copyrighted material without a license as long as the result meets certain criteria, including transforming something into a new, original work – which, AI companies argue, is exactly what they’re doing.
Copyright holders argue that these companies wouldn’t exist at all without their work – as ABBA’s Björn Ulvaeus, who is currently the president of international rights society CISAC, put it in a recent speech: “These tools are extraordinary, but they could not have been built without us.”
Suno recently raised $400 million in its Series D round, putting the company’s valuation at $5.4 billion.
A recent case against Anthropic, the company behind the Claude platform valued at nearly $1 trillion, recently concluded with the company being ordered to pay the largest copyright settlement in American history, $1.5 billion, to a collective of authors. The judge in that case found that Anthropic’s training on their work was fair use – but that the company had pirated that work in order to do so.
Suno allegedly did the same thing, according to a recent report from 404 Media covering a hack of Suno’s internal data, which seemed to indicate that Suno illegally scraped material from YouTube.
Another as-yet-untested avenue of attack for rightsholders around fair use as regards AI centers on whether the unlicensed use of a work hurts the current or future market for the original – what’s called “market dilution.” If it is proven that Suno has had a material impact on the overall streaming market for music, it could result in something of an existential crisis for nascent, creative-generative platforms.
Still, some argue that the genie is out of the bottle. “Rather than getting upset about the potentially lost cause of training data, maybe this is a time to think about what you want the world to look like, factoring we will live with increasingly powerful AI models at our fingertips, should we choose to use them,” the artist and researcher Matt Dryhurst wrote in a recent essay on music being used in AI training.
“Music doesn’t stop. Neither do we,” writes Suno on its website.
