Germany

[DE] Collecting society files model lawsuit against OpenAI to clarify remuneration rights and usage exemptions

IRIS 2025-1:1/24

Christina Etteldorf

Institute of European Media Law

In mid-November 2024, the Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte (German collecting society for music rights – GEMA) announced that it was the first collecting society in Europe to file a lawsuit against the US company OpenAI for the unlicensed use of protected musical works. GEMA wants the courts to clarify the remuneration obligation that it thinks should arise from OpenAI’s systematic use of its repertoire to train its AI systems. It has also asked the Landgericht München (Munich District Court) to decide whether the public communication of original song lyrics by ChatGPT is lawful.

GEMA is a collecting society for musical works that manages the copyright-protected exploitation rights of around 90 000 members in Germany and around 2 million rights holders worldwide. Its primary role is to grant exploitation rights in return for a fee that is then passed on to the relevant rights holder. The lawsuit it filed against OpenAI Ireland Ltd. (the operator of the ChatGPT AI tool in Europe) and its American parent company OpenAI L.L.C. was based on the claim that the rights holders’ works were being used in contravention of copyright. Despite generating billions of dollars in turnover, the company had not acquired any licences or paid any usage fees. The works were being used to train AI systems and original song lyrics were being communicated to the public by the company’s chatbot. GEMA assumed that song lyrics from its repertoire were being used because, firstly, ChatGPT could reproduce them at its users’ request, which was a case of communication to the public under copyright law. Secondly, this was only possible if the AI system on which the chatbot was based had been fed the relevant lyrics in order to be trained. GEMA claimed in its press release that this amounted to deliberate copyright infringement through systematic use of third-party works.

The Munich District Court will now examine various questions concerning a range of different legal aspects. In particular, it will need to decide whether training AI systems with and communicating protected song lyrics are activities covered by copyright law and what the consequences are for the act of communication if the lyrics are freely accessible on the Internet. If it decides that these activities are covered by copyright law, it may need to address exemptions that could be cited by OpenAI. These particularly include text and data mining, which is regulated in Article 44b of the German Urheberrechtsgesetz (Copyright Act – UrhG). According to this provision, the reproduction of lawfully accessible works is permitted in order to carry out automated Internet searches and gathering of information for analysis purposes. In another case at the end of September 2024, the Landgericht Hamburg (Hamburg District Court) decided that Article 44b UrhG could be used to justify the training of AI systems (Judgment of 27 September 2024, case No. 310 O 227/23). However, this case concerned text and data mining for scientific purposes, which is governed by Article 60d UrhG, implementing Article 3 of Directive (EU) 2019/790. Nevertheless, since commercial text and data mining may also fall under Article 44b UrhG, it may not be permitted if the rights holder has made an effective reservation of use. A reservation of use in the case of works that are available online is only effective if it is made “in a machine-readable format”. The dispute with OpenAI is also expected to revolve around whether GEMA has made such an effective reservation of use. The Hamburg District Court had decided – without needing to refer to Article 44b UrhG because of the relevance of Article 60d UrhG – that a declaration in natural language was sufficient to achieve this because AI was able to understand it.