A Munich court has ruled that large language models, including ChatGPT, infringed German authors’ rights by using song lyrics in responses without paying licence fees. Judge Elke Schwager of Munich District Court I found OpenAI liable for damages for the unauthorised use of lyrics, though she did not set a sum; the decision can be appealed.
The action was brought by GEMA, the German collecting society representing authors. GEMA cited nine specific songs as examples, among them Herbert Grönemeyer’s “Männer (Men)”, Reinhard Mey’s “In der Weihnachtsbäckerei (In the Christmas bakery)”, and “Atemlos (Breathless)”, associated with Kristina Bach and Helene Fischer. GEMA’s lawyer said the ruling could be significant across Europe because the relevant rules are harmonised, and he expects negotiations with companies like OpenAI over appropriate licensing fees, stressing the aim of compensation rather than removal from the market.
The court highlighted the particularities of German authors’ rights (Urheberrecht), which emphasize the individual creator and include inalienable rights that differ from Anglo-American copyright approaches that treat rights more as transferable property. Judge Schwager said she was surprised OpenAI had not addressed what she described as a clear legal situation, and concluded that using third-party content to build a product requires payment or permission. The judge characterized the model’s use of lyrics as unlicensed reproduction and distribution and defended authors’ rights as protected intellectual property.
OpenAI said it disagrees with the ruling and is considering next steps, reiterating respect for intellectual property and noting ongoing talks with rights organisations globally. During the trial both sides accepted that lyrics had been used in training the fourth iteration of ChatGPT; the dispute focused on whether the model actively stores and reproduces specific texts. OpenAI argued that the model does not store exact training items but encodes patterns in its parameters and that outputs are generated in response to user prompts. The court rejected the suggestion that reproduction of long, complex song texts could plausibly occur by chance, citing its press statement that coincidence could be ruled out.
The decision drew support from parts of the journalism sector. The head of the German Journalists’ Union described the ruling as a partial victory for authors and said it strengthens the legal position for seeking compensation from AI companies, echoing wider concerns about the use of journalistic and creative content in AI training. GEMA previously pursued restrictive measures for German music on YouTube before reaching a later agreement to allow publication on the platform.