GEMA, Germany's music performing rights society, sued OpenAI after ChatGPT reproduced copyrighted song lyrics in its outputs. The Munich District Court ruled that ChatGPT's verbatim reproduction of lyrics constitutes unauthorized reproduction under German copyright law — the first European court to find AI output directly infringing.
Holding
The Munich District Court held that when ChatGPT outputs verbatim or near-verbatim reproductions of copyrighted song lyrics in response to user prompts, this constitutes unauthorized reproduction under §16 of the German Copyright Act (UrhG). The court ordered OpenAI to implement technical measures to prevent such output.
Arguments For / Positive Implications
- First European ruling directly addressing infringing AI output (not just training)
- Protects songwriters and composers from uncompensated reproduction of their creative works
- Forces AI companies to implement output filters and guardrails for copyrighted content
- Strengthens the position of collecting societies (CMOs) in AI licensing negotiations
Arguments Against / Concerns
- Places the burden of content filtering on AI companies, which is technically challenging at scale
- May lead to over-blocking of legitimate uses (e.g., quoting lyrics for commentary or criticism)
- Does not address whether AI training itself infringes copyright, only output-side liability
- Limited to German jurisdiction — other EU member states may reach different conclusions
Our Takes
This case perfectly illustrates why the AI copyright debate needs to distinguish between training and output. The Munich court didn't rule on whether training on lyrics is lawful — it ruled that reproducing them verbatim in output is unlawful. That's actually a straightforward application of existing copyright law: if the AI quotes a song word-for-word, someone is making a copy. The harder question — who is liable, and how to prevent it — is where things get complicated.Lawra (The Moderate)
This is the most common-sense ruling in the entire AI copyright landscape. ChatGPT was spitting out complete song lyrics on demand — that's reproduction, plain and simple. It doesn't matter that a machine did it instead of a human. GEMA's members wrote those lyrics and deserve to be compensated when they're reproduced. Every AI company that has been hoping the copyright problem would go away just got a European wake-up call.Lawrena (The Skeptic)
The ruling is technically correct but practically limited. Yes, verbatim reproduction of lyrics is infringement — no one seriously disputes that. The real value of this case is that it forces OpenAI and other AI companies to build better output filters. That's actually good for the industry: responsible companies will implement guardrails, and the technology to detect and prevent copyrighted output will improve. This is a solvable engineering problem, not an existential threat.Lawrelai (The Enthusiast)
The distinction between training and output is critical, and this court got it right. Learning from lyrics to understand language patterns is one thing; reproducing them verbatim is another entirely. The spirit of copyright is to incentivize creation — and outputting exact copies of someone's creative work with no attribution or compensation violates that spirit directly. But the solution is engineering guardrails and licensing agreements, not banning AI from learning from musical works. GEMA and OpenAI should be negotiating collective licensing frameworks that compensate creators while enabling innovation.Carlos Miranda Levy (The Curator)
Why This Case Matters
GEMA v. OpenAI is the first European court ruling to find that an AI system’s output directly infringes copyright. While most AI copyright litigation has focused on whether training on copyrighted works is lawful, this case addresses the other side of the equation: what happens when the AI’s output reproduces copyrighted material?
What Happened
GEMA (Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte) is Germany’s equivalent of ASCAP or BMI — a collective management organization that represents songwriters, composers, and music publishers. GEMA manages the rights to over 1 million musical works in Germany.
GEMA discovered that ChatGPT would, in response to certain prompts, output complete or near-complete lyrics of copyrighted songs managed by GEMA. This included lyrics to works by major German and international artists. GEMA filed suit in the Munich District Court, alleging unauthorized reproduction under §16 of the German Copyright Act.
The Court’s Reasoning
The Munich court applied a straightforward analysis:
-
Reproduction: When ChatGPT outputs copyrighted song lyrics verbatim, it creates a reproduction within the meaning of §16 UrhG. The court rejected OpenAI’s argument that the output is a “new creation” by the model — the lyrics are objectively identical to the original works.
-
Liability: OpenAI, as the operator of ChatGPT, bears responsibility for the system’s output. The court did not accept the argument that users bear sole responsibility for prompting the system to produce infringing content.
-
No applicable exception: German copyright law’s exceptions for text and data mining (§44b UrhG) apply to the training process, not to the output. The court found no statutory exception that would permit the unauthorized reproduction of song lyrics in AI output.
Training vs. Output
This case is notable for what it does not decide. The court expressly limited its ruling to output-side infringement — the reproduction of copyrighted lyrics in ChatGPT’s responses. It did not rule on whether OpenAI’s training on copyrighted lyrics was itself an infringement. That question remains open in German law and will likely be addressed in future litigation.
The Broader Impact
GEMA v. OpenAI establishes that AI companies can be held liable for copyright-infringing output under European law. This has immediate practical implications: AI companies will need to implement more sophisticated content filters to prevent the reproduction of copyrighted material in their outputs. The ruling also strengthens the negotiating position of collecting societies seeking licensing agreements with AI companies for the use of their members’ works.
Sources
- GEMA v. OpenAI, Az. 42 O 2915/24 (LG München I, Nov. 7, 2025) (2025-11-07)
- German Court Finds ChatGPT Lyrics Output Infringes Copyright — Reuters (2025-11-07)
- GEMA Wins Landmark AI Copyright Case Against OpenAI — GEMA Press Release (2025-11-07)
Explore Legal Frameworks
Cases don't happen in a vacuum. Explore the regulatory frameworks shaping AI law around the world — from the EU AI Act to emerging legislation in Latin America.
Ready for structured learning? Explore the Learning Program →
Comments
Loading comments...