Introduction
The media and entertainment industry is changing rapidly with the increasing use of artificial intelligence. From creating music to generating scripts, AI systems are now increasingly intertwined with artistic creation. Nonetheless, these technological advancements have also raised significant legal concerns, particularly regarding the use of copyrighted content in AI training. A major development in this field was the case of GEMA v. OpenAI, in which the Regional Court of Munich (Landgericht München I) delivered its decision in November 2025. The case raised an important question: can AI companies train their models on copyrighted song lyrics without the knowledge or consent of rights holders? GEMA, a collective management organisation representing composers and music publishers in Germany, claimed that OpenAI had used copyrighted song lyrics without obtaining a licence. GEMA argued that the chatbot was able to recreate sections of popular songs when prompted, which constituted copyright infringement. The Munich court accepted this argument, and GEMA prevailed. The decision is widely considered to be one of the first major court rulings directly addressing the intersection of generative AI and copyright law. It has far-reaching consequences not only for technology companies, but also for musicians, publishers, and other creative professionals worldwide.
Background of the Dispute
GEMA is one of the most influential copyright collecting societies in Europe. It manages the rights of thousands of composers, lyricists, and music publishers. The organisation ensures that artists are compensated when their works are performed, reproduced, or distributed. The dispute began when GEMA discovered that ChatGPT could generate the lyrics of several well-known German songs when prompted by users with only a few words. In certain instances, the generated lyrics were almost identical to the original compositions. GEMA alleged that these findings demonstrated that the AI model had been trained on copyrighted lyrics without permission from the authors or publishers. The organisation filed a lawsuit against OpenAI in the Munich Regional Court, claiming that this use violated the exclusive rights of reproduction and communication to the public granted to copyright holders. The case particularly concerned several popular German songs written by well-known composers and lyricists represented by GEMA. According to the claim, these works were part of the data used to train the language model. OpenAI defended itself by arguing that its system does not store complete songs or texts. Instead, the company stated that large language models analyse patterns from extremely large datasets. The model is not designed to store exact copies of works but instead learns statistical relationships between words. The company further argued that if a user generated song lyrics through a prompt, responsibility would lie with the user rather than the developer of the AI system.
Key Legal Issues
The court had to consider several legal questions. The first issue was whether training an AI model on copyrighted material constitutes “reproduction” under copyright law. Copyright protection grants authors the exclusive right to reproduce their works in any form. Therefore, if the AI system created copies of lyrics during the training process, it could potentially constitute unlawful reproduction. The second issue concerned the outputs of the chatbot. If the AI system was able to generate song lyrics or close approximations when prompted, the court had to determine whether this amounted to unauthorised communication of the work to the public. Lastly, the court examined whether any legal exceptions permitted the use of copyrighted material. In particular, OpenAI relied on the text-and-data mining exception, which allows certain forms of copying for analytical purposes.
The Court's Reasoning
After analysing the evidence and the parties’ arguments, the Munich Regional Court concluded that copyright infringement had occurred. One of the most significant considerations in the judgment was the possibility that AI models can memorise parts of their training data. Although AI developers often argue that models only learn statistical patterns, the court relied on research suggesting that large language models may reproduce specific portions of training data. Because ChatGPT was able to generate lyrics that were very similar to original songs, the court concluded that the model must have been trained on such lyrics. The judges therefore found that the training process involved copying copyrighted material into the AI system. Under German and European copyright law, reproduction does not require a traditional physical copy, such as a printed page. Any digital reproduction of a work may qualify as reproduction. The court also examined the chatbot’s responses to user prompts. It determined that when the AI generated song lyrics and presented them to users, this amounted to making the works accessible to the public. Since the rights holders had not authorised such use, this conduct interfered with their exclusive rights. The text-and-data mining exception was another important aspect of the case. While this exception allows limited copying for analytical purposes, the court held that it did not apply here. The judges reasoned that text-and-data mining is intended to analyse information, not to incorporate works into a system capable of reproducing them later. Because the lyrics were effectively embedded within the model and could be recreated through prompts, the exception could not be relied upon as a defence. Finally, the court rejected the argument that users should bear responsibility for the outputs. It held that the developer and operator of the AI system ultimately control the training process and model design, and therefore bear responsibility for the resulting outputs.
The Court's Decision
Based on its analysis, the Munich Regional Court ruled in favour of GEMA. The court found that OpenAI had infringed the copyrights of songwriters and publishers represented by GEMA. It held that both the training of the model and the generation of lyrics through the chatbot interfered with the exclusive rights of the authors. As a result, OpenAI was ordered to cease the unlawful use of the lyrics and disclose information regarding the use of the works. The company may also face liability for damages depending on future proceedings. Although the case concerned a specific set of songs, the reasoning of the judgment has broader implications for the generative AI industry as a whole.
Consequences for the Media and Entertainment Industry
The ruling has attracted worldwide attention because it directly addresses the relationship between AI technology and copyright law. For creators and rights holders, the decision represents an important step toward protecting artistic works. Many artists have expressed concern that their works could be exploited by artificial intelligence systems without credit or compensation. The Munich ruling reinforces the idea that copyright law remains a powerful tool for protecting creative works in the age of artificial intelligence. For technology companies, the decision highlights the legal risks associated with training AI models on copyrighted data without licences. Developers may need to reconsider how training datasets are collected and used. Many experts believe that licensing agreements between AI companies and content creators will become increasingly common. The case also demonstrates that different jurisdictions may adopt different approaches to AI and copyright. Courts in other countries are addressing similar issues and, in some instances, reaching different conclusions. This creates a complex international legal environment for companies developing generative AI technologies. In the long term, the case may encourage lawmakers to develop clearer rules governing the use of copyrighted material in AI training. Governments and regulators are already debating how intellectual property law should evolve alongside rapidly advancing technologies.
Conclusion
The GEMA v. OpenAI decision marks an important milestone in the evolving relationship between artificial intelligence and copyright law. By ruling that AI training and outputs may infringe copyright where song lyrics from identifiable works are reproduced, the Munich Regional Court has set an important precedent for future cases. The judgment reinforces the principle that technological innovation does not exist outside the scope of intellectual property law. Even advanced AI systems must respect the rights of creators whose works contribute to the training of such technologies. For the media and entertainment industry, the case signals that copyright protections continue to apply in the digital and AI-driven era. Artists and rights holders are likely to play a significant role in shaping how AI technologies develop and operate. As generative AI continues to evolve, courts, policymakers, and industry stakeholders will need to balance technological innovation with the protection of creators’ rights. The GEMA v. OpenAI ruling represents one of the earliest and most significant steps in that ongoing process.
References
- Munich Regional Court (Landgericht München I), Press Release on the Judgment in GEMA v. OpenAI, 11 November 2025.
- Von Lewinski, Silke. GEMA v. OpenAI: The Munich Regional Court’s Approach to Generative AI and Copyright. GRUR International, 2026.
- Reuters. German Court Says OpenAI Used Song Lyrics in Violation of Copyright Law. November 2025.
- Reuters. Getty Images Largely Loses Landmark UK Lawsuit Over AI Image Generator. November 2025.
- Getty Images v. Stability AI Ltd., High Court of Justice (England and Wales), 2025.