📜 1. Case Summary: GEMA vs. OpenAI
The Case of the Memorised Lyrics
In a landmark decision, the Munich Regional Court ruled in favor of GEMA, Germany's music collecting society. GEMA argued that OpenAI used the lyrics of famous songs like "Atemlos" to train its GPT-4 models without obtaining a licence. They proved that the AI could reproduce the lyrics almost verbatim when prompted.
OpenAI argued that their models do not store copies but only learn statistical correlations. They claimed this falls under the TDM (Text and Data Mining) exception for scientific research. However, the court found that the "hallucinations" or minor errors in the output did not change the fact that the original work was recognizable.
The court ruled that the memorisation of the lyrics in the model's parameters constituted an unauthorized fixation (or reproduction). As a result, the court granted an injunction, ordering OpenAI to stop using the lyrics, and opened the door for potential damages.
🔗 2. Legal Terminology Match
Match the legal term to its definition. Correct pairs will disappear and reappear at the end!
⚖️ 3. Understanding the Verdict
Select the correct interpretation based on the Munich ruling.
1. Why did the court reject OpenAI's "TDM" defense?
2. What did the court decide regarding the "hallucinations" (errors) in the lyrics?
📝 4. Legal Context Gap Fill
Complete the summary of the case logic.
Word Bank
1. GEMA successfully applied for an ____ to stop the use of the lyrics.
2. The court found that the ____ of training data occurred in the model parameters.
3. OpenAI claimed the model only learns ____ correlations, not actual text.
4. Storing the lyrics in the model weights counts as a ____ of the work.
5. To use protected songs legally, AI companies usually need a ____.
🎉 Case Closed!
You have analyzed the GEMA vs. OpenAI ruling. Review terms below:
- Collecting Society
- Injunction
- Fixation
- Memorisation
- Reproduction
- Licence
- TDM
- Liability