Case Summary: The plaintiff is the German Music Copyright Association (GEMA), a collective management organization representing composers, lyricists, and music publishers in exercising their music copyrights. The defendant is the American OPENAI group, which develops and applies chatbots.
The plaintiff accuses the defendant of using the lyrics of nine songs, which it represents, to train its language model without permission, and of outputting these lyrics in whole or in part to the public through the chatbot, constituting an infringement of copyright and moral rights of the authors.
Key Points of the Ruling:
The Munich District Court delivered its ruling on November 11, 2025.
1. “Memorization” as “Copying”: The court determined that the behavior of the AI model in memorizing copyrighted lyrics text in its internal parameters during training constitutes a “copying” act as defined in Section 16 of the German Copyright Act (UrhG). The court found that these texts are “physically fixed” in the model and can be perceived again through user prompts.
2. Chatbot Output Constitutes Infringement: The act of the chatbot outputting the disputed lyrics to users was recognized as a prohibited act of “making available to the public” (public dissemination) under Section 19a of the Copyright Act. Outputting modified lyrics constitutes a prohibited act of “adaptation” under Section 23.
3. Furthermore, the court rejected the defendant’s defenses regarding “exceptions for scientific research” and “exceptions for text and data mining.”
Commentary:
This ruling emphasizes the necessity of strict compliance with copyright law during the model training phase. AI companies must conduct due diligence on the sources of their training data to ensure they have legal authorization, especially when dealing with copyrighted materials. Data that explicitly states rights are reserved (such as through robots.txt or website terms) must be excluded from the training set.
The ruling has provided relief to rights holders, indicating that the existing copyright law framework can effectively address the challenges posed by generative AI.
The analysis of copyright issues related to training data for generative AI in this ruling is thorough and in-depth, and its reasoning may have far-reaching implications for legislation and judicial practices regarding AI and intellectual property in the EU and globally.
This case can be compared with the previous issue of the “Getty Case in the UK: AI Model Does Not Infringe Copyright.”
This article is rewritten based on a summary of the 200-page ruling on AI.