The Munich Regional Court, specializing in copyright law, granted claims for injunctive relief, disclosure and damages asserted by GEMA against two companies of the Open AI group.
The ruling concerned the lyrics of nine well-known German songwriters (including “Atemlos” by Kristina Bach and “Wie schön, dass du geboren bist” by Rolf Zuckowski).
The Plaintiff, GEMA (Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte) is the German copyright body for music authors and publishers. It argued that the song lyrics were memorized in the defendant’s language models and, when the chatbot was used, were largely faithful to the original as answers (outputs) to simple user queries.
OpenAI operates language models and chatbots based on them. They objected to the claims, arguing that their language models did not store or copy specific training data, but rather reflected in their parameters what they had learned based on the entire training dataset. The two OpenAI companies were not identified in the Court’s press release.
The defendant’s view
OpenAI contended that, because the outputs were generated solely as a consequence of user input (prompts), that the respective user, as the producer of the output, was responsible for it, not OpenAI. In any case, OpenAI asserted that any potential infringements were covered by the limitations of copyright law, in particular the limitation for so-called text and data mining.
The Court’s view
The court found that the song lyrics in question were reproducibly contained in the defendant’s language models 4 and 40. Information technology research has established that training data can be contained in language models and extracted as outputs. This is referred to as memorization. Such memorization occurs when, during training, the language models not only extract information from the training dataset.
This memorization was confirmed by comparing the song lyrics contained in the training data with their reproduction in the outputs. Given the complexity and length of the song lyrics, chance as the cause of their reproduction is ruled out.
Copyright law does apply
New technologies such as language models are covered by the reproduction right under the German Copyright Act, Article 2 of the InfoSoc Directive and Section 16. According to the case law of the Court of Justice of the European Union, indirect perceptibility is sufficient for reproduction, which is given when the work can be perceived with the aid of technical means.
According to the court’s decision, the defendants also unlawfully reproduced and made publicly available the song lyrics in question by displaying them in the chatbot’s output. The original elements of the lyrics would always be recognizable in the output.
Why it matters
The defendants, not the users, are responsible for this. The outputs were generated by simple prompts. The defendants operated the language models, for which the song lyrics were selected as training data and with which they were trained. They are responsible for the architecture of the models and the memorization of the training data. Thus, the language models operated by the defendants significantly influenced the outputs; the specific content of the outputs is generated by the language models.
Further reading
GEMA ruling against Open AI (Case No. 42 O 14139/24). Press release. November 11, 2025. Cornelia Kallert, Presiding Judge at the Munich I Regional Court
