On November 11, 2025, the Munich District Court I delivered the first-instance judgment in the case filed by GEMA (the German Society for Musical Performing and Mechanical Reproduction Rights) against OpenAI (Case No. 42 O 14139/24). The court held that OpenAI infringed copyright by using copyrighted song lyrics without authorization to train its ChatGPT model. This landmark ruling represents the first judicial evaluation in Europe of generative AI training on copyrighted works and has significant implications for global AI-related copyright disputes.

I. Background and Key Issues

1. Origin of the Case 

GEMA is one of Europe’s largest collective management organizations, representing about 100,000 songwriters and music publishers. In November 2024, GEMA filed a lawsuit accusing OpenAI of using the full lyrics of nine popular German songs without permission for training ChatGPT. GEMA argued that OpenAI failed to obtain licenses, pay royalties, or comply with copyright restrictions, and that ChatGPT could regenerate lyrics through its “memory-reproduction” mechanism. GEMA contended that such use did not fall under text and data mining (TDM) exceptions.

2. Main Issues in Dispute 

• Whether OpenAI’s use constitutes copyright-relevant “reproduction” and “making available to the public” 

• Whether OpenAI can rely on the TDM exception under Article 44b of the German Copyright Act 

• Whether GEMA has proper standing 

• Whether prohibiting the infringing activity violates the principle of proportionality 

II. Court’s Findings

1. Copyright Infringement 

The court found that OpenAI reproduced and made copyrighted lyrics available to the public. 

• Reproduction: Model parameters contained fixed, reproducible lyric information accessible through simple prompts, satisfying the “fixed + perceptible” requirement under Article 16. 

• Making available: ChatGPT outputs lyrics to the general public, meeting Article 19a’s definition of “making available.”

2. TDM Exception Not Applicable 

The court held that: 

• Article 44b applies only to temporary reproductions necessary for data analysis. 

• OpenAI’s long-term storage and regeneration of lyrics exceeded the scope of TDM. 

• GEMA had issued a valid machine-readable TDM opt-out, and OpenAI failed to conduct reasonable checks.

3. Standing and Proportionality 

The court held: 

• GEMA had acquired exclusive rights and enforcement authority. 

• Prohibiting infringement did not unduly harm OpenAI’s interests; OpenAI could obtain licenses.

4. OpenAI Not a “Research Institution” 

OpenAI operates commercially; thus, research-related copyright exceptions did not apply.

III. Judgment and Impact

1. Judgment 

The court ordered: 

• OpenAI must cease copying and providing lyrics through ChatGPT under penalty of up to €250,000 or six months detention. 

• OpenAI must disclose the scope of infringement and associated profits. 

• OpenAI must compensate GEMA for all losses (amount to be determined). 

• GEMA may publish the judgment in national newspapers at OpenAI’s expense. 

• OpenAI must pay €4,620.70 in legal fees plus statutory interest.

2. Significance 

• Marks a new era in Europe for AI copyright oversight. 

• Sets a global reference for resolving AI copyright cases. 

• Signals increasing regulatory pressure on AI companies to obtain proper licenses.