A regional court in Munich has ordered OpenAI to pay damages to GEMA, Germany’s leading music rights society, after finding that ChatGPT reproduced lyrics from German songs during user queries.
The ruling followed evidence presented by GEMA that OpenAI’s model generated passages closely matching copyrighted works by several well-known artists, including Herbert Grönemeyer.
Reuters reported that the damages amount was not disclosed publicly, although the court confirmed OpenAI is financially liable.
GEMA argued that the appearance of these lyrics was only possible because the underlying material had been ingested during training without a licence.
The organisation represents roughly one hundred thousand songwriters, composers and publishers, giving it one of the largest repertoires of licensed content in Europe.
OpenAI contested the claim, stating that the model does not store material in a way that allows wholesale reproduction and that any overlaps resulted from statistical processes rather than direct retention.
The company added that it believes the court’s interpretation does not reflect how large language models operate and indicated it is considering an appeal.
Europe’s Copyright Framework Meets Generative AI
The ruling arrives at a moment when European institutions are paying closer attention to how AI developers source their training data. Many rights-holders, including publishers, news agencies and music organisations, have argued that models trained on unlicensed content effectively benefit from creative work without permission.
For GEMA, the case is part of a broader campaign to establish that artistic material, including lyrics, should be licensed before being used to develop commercial AI products. The organisation has pursued similar conversations with technology companies in recent months, calling for transparent data-handling policies and compensation mechanisms.
Legal specialists told several German outlets that the Munich judgment may not be the final word on this issue, but it highlights growing judicial willingness to treat unlicensed training data as a form of copyright use rather than incidental activity.
If upheld through appeals, the order could influence how companies negotiate licences or reshape their training datasets to avoid future disputes.
Ruling Could Influence The Industry
The OpenAI case is being watched across Europe because it touches on fundamental questions about how large language models learn. If courts begin to treat training data as requiring the same clearance as published reproductions, companies may face higher costs, slower development cycles or the need to secure broad agreements with music, media and publishing groups.
For creators, the decision offers a sense of momentum in ongoing efforts to make AI developers more accountable. For the technology sector, it introduces renewed pressure to build systems that respect intellectual property while still maintaining access to the diversity of material required to train high-performance models.
OpenAI now faces a choice between securing licences, adjusting model behaviour or challenging the ruling through Germany’s appeals system. As the case progresses, it will shape the conversation about what fair use means in the age of generative intelligence and how cultural assets are treated when machines learn from human creativity.
