It’s a business model and it’s looking good. But not for AI.
In a world-first ruling, the Munich Regional Court found OpenAI guilty of copyright infringement for training ChatGPT on German song lyrics without permission.
This is not just another legal hiccup. this week’s decision marks the first time any court worldwide has ruled against OpenAI in a copyright case, and the implications could fundamentally reshape how AI operates globally.
German collecting society GEMA did not just win a symbolic victory, they showed that ChatGPT systematically memorized and reproduced copyrighted material from nine German songs without paying a single licensing fee. The ruling found that both the storage of copyrighted material in AI models and its reproduction in chatbot responses constitute clear violations of copyright law.
GEMA’s CEO Dr. Tobias Holzmüller delivered a pointed message that should make every AI executive sit up: The internet is not some kind of “self-service buffet”, and creative achievements by human beings are not simply templates for use free of charge.
Smoking gun
The evidence against OpenAI was damning. During the trial, it was undisputed that ChatGPT uses artists’ lyrics for training purposes, court documents revealed. GEMA demonstrated that ChatGPT could reproduce protected lyrics by well-known authors in response to simple prompts, proof that the AI had memorized copyrighted content word for word.
The court examined specific songs including “Männer (Men)” by Herbert Grönemeyer and “Atemlos (Breathless)” by Kristina Bach. Judge Elke Schwager rejected OpenAI’s defense, finding that coincidental generation of text matching lyrics exactly or substantially was not plausible. The judges agreed that both memorization in language models and reproduction of song lyrics in chatbot outputs constitute copyright infringement, legal experts noted following the decision.
The court rejected OpenAI’s claim that it qualifies as a privileged research organization and ruled that legal authorization for text and data mining does not justify storing and outputting copyright-protected material.
Multi-billion dollar reckoning
This ruling is not just about OpenAI. The court’s decision is being hailed as a landmark ruling and significant setback for operators of generative AI systems worldwide.
The financial fallout is just beginning. OpenAI must now obtain GEMA licenses and pay damages for unauthorized usage, though the exact figure has not been disclosed.
GEMA has another lawsuit pending against AI music generator Suno, scheduled for January 26, 2026, and represents over 100,000 members in Germany plus two million rights holders worldwide.
While the ruling only applies to older versions of ChatGPT 4 and 4o, it sets a legal framework that could force every AI company operating in Europe to rethink their entire approach to training data.
The age of free-for-all AI development may be over.
Morgan Freeman didn’t mince words: when AI clones your iconic baritone without permission, you call it robbery.