OpenAI may not use lyrics without license, German court rules (www.reuters.com)

A Munich court ruled that OpenAI may not use song lyrics without a licence, siding with German music rights society GEMA in a landmark copyright case. Judge Elke Schwager ordered OpenAI to pay damages after GEMA argued that ChatGPT reproduces copyrighted German lyrics and that the model was trained on protected material from the society’s roughly 100,000 members. OpenAI countered that GEMA misunderstood how ChatGPT works; both parties said they would comment further and the decision can be appealed. The ruling matters for the AI/ML community because it directly targets both the use of copyrighted material in training sets and the model’s naked outputs. If upheld, it could force developers to secure licences for copyrighted text used in training and/or constrain model output that reproduces such works, changing dataset collection, filtering, and legal risk models. Practically this may accelerate adoption of licensed corpora, stronger provenance tracking, output filters, and “safe” training practices (or costly licensing arrangements), and could alter commercial viability for smaller teams. The case is likely to be cited across Europe as regulators and rights holders push for frameworks that require payment for musical works in both training and generated outputs.