AI copyright news, graded by source

Munich court rules that ChatGPT's memorization of nine German song lyrics in its models, and their reproduction in outputs, infringes copyright and is not covered by the text-and-data-mining exception

The Munich Regional Court I's copyright chamber ruled on Nov. 11, 2025 that GEMA, the German music collecting society, is entitled to injunctive relief, information and damages from two OpenAI companies over the lyrics of nine well-known German works, including Atemlos and Wie schoen, dass du geboren bist. A personality-rights claim over misattributed altered lyrics was dismissed. Per the court's press release, OpenAI argued its models store no specific training data, that users rather than OpenAI are responsible for outputs generated from prompts, and that any use was covered by the text-and-data-mining exception. The court found the lyrics were reproducibly contained in the GPT-4 and GPT-4o models, shown by comparing the lyrics in the training data with the outputs, with chance ruled out given their length; that this memorization is a reproduction under Section 16 UrhG and the InfoSoc Directive even though fixed in probability values; that the TDM exception in Section 44b UrhG covers preparatory copies made to assemble a training corpus but not reproduction of works inside the model; and that outputs from simple prompts are OpenAI's responsibility. The judgment is not final.