Berlin — A German court has ruled that OpenAI must pay licensing fees to GEMA for using copyrighted song lyrics in its AI models, including ChatGPT. Judge Elke Schwager found that the lyrics were memorized in training data and reproduced verbatim when prompted. OpenAI plans to appeal and argues models learn patterns, not verbatim texts.
The court ordered compensation for all damages, including unpaid royalties and legal costs, plus interest. It held that the defendants — not users — are responsible for outputs produced by their language models, underscoring a direct link between training data and generated content. The case concerns lyrics from nine German songwriters represented by GEMA.
GEMA has argued for fair remuneration as AI becomes more embedded in daily life. The ruling suggests potential payments running into hundreds of thousands of euros, though OpenAI has not admitted liability and may pursue further legal avenues. The decision emphasises that creators should be compensated for use in training data.
This could set a precedent for how European copyright law intersects with AI training, especially as GEMA operates a licensing model since 2024 and a parallel SunoAI case awaits a hearing next year. Critics warn the decision could hinder innovation or restrict access to AI tools, while supporters argue it protects artists and aligns incentives for creators in a fast-evolving tech landscape.
