Via German tech law firm Vossius we have learned that the Munich Regional Court ruled in favor of GEMA, Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte, the German counterpart to RIAA in the USA, in a case about nine songs that were the subject of copyright infringement.
The core of the case is ChatGPT which upon request was able to reproduce chorus and the first two verses from the songs in question.
The court ruled that the way that LLMs, Large Language Models work on a basic level can be classified as reproduction in relation to copyright within the legal frame work of the European InfoSoc Directive. The court also found that the data mining used by generative AI was not exempt under existing rules, as generative AI still uses and sometimes outright reproduces the data upon which it is trained.
This is a direct violation on the European Digital Single Market directive, which states that such use requires a license from the original creator. OpenAI counter-argued that the latest version, GPT5 have build-in counter measures to avoid direct reproduction, but the court pointed out that the output provided by ChatGPT still constitutes public reproduction, and that the selection of training data plays a pivotal role.
Both RIAA in the US and KODA in Denmark are currently engaged in identical lawsuits against not only OpenAI, but also SUNO and similar services, with court dates not being official yet.