Florian Mueller of AI Fray provides an excellent summary of the trial held by Landgericht München I (Munich I Regional Court) in the lawsuit filed by GEMA, the collecting society representing authors of musical work, against OpenAI.
According to Mueller, “The presiding judge [Presiding Judge Dr. Elke Schwager] left no doubt in her introductory outline of the issues in the case that the panel deems OpenAI liable for copyright infringement.”
According to Dr. Marcus von Welser of VOSSIUS, the court made a preliminary assessment that (i) the processes involved in a LLM can constitute a “reproduction,” and (ii) the text and data mining exception in the EU can also cover the activity of the LLM presumably in deep learning. But the TDM exception does not apply when the LLM has memorized and can reproduce copyrighted training data.
Even though OpenAI trained its model in the U.S., not Germany, the outputs of infringing song lyrics occurred in Germany. The court reportedly was skeptical that OpenAI can escape liability by shifting the responsibility to users of ChatGPT who seek and obtain the outputs of song lyrics.
I’m not sure which song lyrics are in the evidence provided by GEMA. But 9 songs are involved.
The court said it would deliver the decision on November 11, 2025.