A Munich court has ruled that large language models such as ChatGPT, which used song lyrics in their responses without paying licence fees, infringe German authors’ rights. Judge Elke Schwager at Munich District Court I said OpenAI would be liable for damages for the unauthorised use, though she did not specify an amount. The verdict can be appealed.
The case was brought by GEMA, the German association that defends authors’ rights. GEMA used nine specific songs as examples, including “Männer (Men)” by Herbert Grönemeyer, “In der Weihnachtsbäckerei (In the Christmas bakery)” by Reinhard Mey, and “Atemlos (Breathless)” originally by Kristina Bach and popularized by Helene Fischer. GEMA’s lawyer Kai Welp said the ruling could be groundbreaking for Europe because the relevant rules are harmonised and that he expected negotiations with companies like OpenAI on suitable licensing fees. “The goal is not to remove anything from the market, but rather to receive appropriate compensation,” Welp told journalists.
German authors’ rights law (Urheberrecht) differs from Anglo-American copyright: it places more emphasis on the individual author and treats certain rights as non-transferable rather than the property of a content owner such as a publisher or record label. Although the case concerned German law and usage, GEMA argued the decision would have wider implications across Europe.
Judge Schwager said she was astonished OpenAI had not taken heed of what she described as a clear legal situation. “We have highly intelligent defendants who have managed to create the most modern of technologies,” she said. She found that using outside content in creating a product requires payment or permission, concluding the current usage amounted to unlicensed distribution and reproduction. “Authors’ rights are protected intellectual property,” the judge said.
OpenAI said it disagreed with the verdict and was examining further steps, adding that it respected intellectual property rights and was in negotiations with relevant organisations worldwide. During the trial neither side disputed that the songs’ lyrics had been used to train the fourth iteration of ChatGPT. The dispute centered on whether the lyrics were actively stored in the model for future use.
OpenAI argued ChatGPT does not store or copy specific training data but instead reflects in its parameters what it learned from the entire training dataset. It also said outputs are generated in response to user prompts, so responsibility for specific outputs lies more with users than with OpenAI. The court rejected the idea that random generation could plausibly reproduce complex, lengthy song texts by coincidence. “Given the complexity and length of the song text, coincidence can be ruled out as the cause of the reproduction of the song lyrics,” the court wrote in a press release.
The ruling was welcomed by some in the journalism sector. Mika Beuster, chairman of the German Journalists’ Union (DJV), called it “a partial victory for authors’ rights” and said the decision strengthened journalists’ legal position when seeking compensation from AI companies. He added: “The training of AI models is intellectual property theft,” reflecting broader concerns among media organisations about the legality of using journalistic content in AI training.
GEMA previously made headlines with a restrictive approach to German music videos on YouTube, though a deal was later reached to permit their publication on the platform.