Overview
- The regional court in Munich ruled that using GEMA‑registered song texts without licenses to train ChatGPT infringed copyright.
- The case centers on nine German songs, including works by Helene Fischer and Herbert Grönemeyer.
- Judge Elke Schwager held that compensation is due for reproduction within the models and in outputs, with “indirect perception” qualifying as reproduction under EU case law.
- OpenAI was ordered to pay damages and to disclose its training practices and potential revenues to assess indemnity, with the compensation amount undisclosed.
- OpenAI said it disagrees with the decision and is reviewing next steps, while GEMA and observers describe the case as a European first with potential broader impact on dataset transparency and creator remuneration.