The decision

OpenAI condemned in Germany: ChatGPT infringes copyright

According to the Munich court, the San Francisco company used the lyrics of some copyrighted songs and will now have to pay royalties

by Massimo De Laurentiis

11 November 2025

2' min read

Translated by AI

Versione italiana

2' min read

Translated by AI

Versione italiana

The Munich court condemned OpenAI for using copyrighted song lyrics in the training of its ChatGPT language model, in a decision that could become an important precedent for the whole continent.

The court's reasons

OpenAI was sued by GEMA, the German equivalent of the SIAE, which protects the rights of authors, composers and music publishers.

According to the association, the Large Language Model ChatGPT was also trained using the lyrics of some popular German songs and later reproduced them in its outputs without paying any licence or remuneration.

The court ruled that the storage and reproduction of protected material by OpenAI models constitutes copyright infringement. According to the court, in fact, this process is tantamount to the actual physical reproduction of texts, which is sufficient to trigger protection under German law.

The ruling requires the San Francisco company to pay royalties to GEMA, although the amount of damages has not been made public.

OpenAI, sigla accordo da 38 mld usd con Amazon per potenza calcolo

OpenAI's defence

OpenAI defended itself by claiming that ChatGPT does not store or copy specific data, but rather reflects in its parameters what it has learned from the entire training data set. Furthermore, the company tried to shift the responsibility onto the users, stating that OpenAI's templates generate output based on what is entered into the prompts, for which ChatGPT users are to blame.

The court, however, found it 'implausible' that the reproduction of entire tracks or long passages could be accidental, thus rejecting the company's line of defence.

In justifying the decision, Judge Elke Schwager emphasised that a company capable of developing such advanced technology cannot ignore the need to pay a licence fee for content used in model training.

Perspectives and future developments

An appeal is possible and OpenAI has announced further legal action. Experts believe that this case could last several years, even reaching the Court of Justice of the European Union, since the ruling concerns copyright in general.

The German court's verdict contrasts with another recent decision by the High Court in London, which instead dismissed most of the allegations made by Getty Images against Stability AI. In that case, the artificial intelligence company was accused of using copyrighted photos to train Stable Diffusion, its image generation model. During the trial, the main charges were dropped for lack of evidence and the court only recognised partial and limited infringements.

In any case, English and German law are different, and according to a GEMA lawyer, the Munich court decision may set a significant precedent for the whole of Europe because the rules are harmonised.

This ruling could also pave the way for new negotiations with technology companies to define a remuneration system for content used by AI.