Case Commentary - GEMA versus OpenAI: The Key Takeaways

In this case commentary, Mohamed Ataul Karim analyses the Munich Regional Court’s landmark ruling in GEMA v. OpenAI, highlighting the court’s findings on fixation, storage, and memorisation of song lyrics in AI models and its rejection of the text and data mining exception. He outlines why the decision represents a significant development for copyright liability in AI training under German and EU law.

Download

Click here for a summary or to preview text

Introduction

The claimant, GEMA, is a German music collecting society that manages the rights to musical works transferred to it by composers, lyricists, and music publishers. The defendants, OpenAI Ireland Ltd (operator of ChatGPT in Europe) and OpenAI L.L.C. (parent company), are a group of companies that develop, operate, and license generative artificial intelligence. The dispute concerns the use of nine (9) German song lyrics to train GPT large-language models and chatbots (Models 4 and 4o), without obtaining a licence from GEMA. The claimant sued the defendants in the 42^nd Civil Chamber of the Munich I Regional Court for copyright infringement arising from the unauthorised use of those song lyrics (case no. 42 O 14139/24). On November 11, 2025, the Court ruled in favour of the claimant. This paper outlines two key takeaways of this decision.