German Court Rules Against OpenAI on Copyright: How Far Can Learning Song Lyrics Go?

1. "AI Should Not Memorize Lyrics for Free" — The Details of a Historic Ruling

On November 11, 2025, a ruling by the District Court in Munich, Germany, sent ripples through the entire generative AI industry.
The court found OpenAI in violation of German copyright law because its ChatGPT had learned the lyrics of popular German songs without permission and could reproduce them almost verbatim at the user's request.The Guardian

The lawsuit was filed by GEMA, a German copyright management organization representing about 100,000 composers, lyricists, and music publishers. They claimed that at least nine songs, including Herbert Grönemeyer's "Männer" and "Bochum," and Helene Fischer's "Atemlos durch die Nacht," were used for both learning and output despite being protected.The Guardian

The court ruled that

• incorporating lyrics during the model's learning process and retaining them internally ("memory")

• and reproducing the lyrics at the user's request ("output")

both constituted copyright infringement.The Japan Times

OpenAI is required to pay damages to GEMA and license fees for the lyrics in question. Although the exact amount has not been disclosed, some media report that it could reach "hundreds of thousands of euros," including unpaid fees, interest, and litigation costs.euronews

TechCrunch has described this ruling as "the first full-scale AI copyright landmark decision in Europe," and reported that OpenAI "disagrees with the ruling but is considering its response."TechCrunch

2. GEMA's Message: "The Internet is Not a Self-Service Store"

A widely circulated comment from GEMA CEO Tobias Holzmüller emerged from this ruling.

"The internet is not a self-service store, and human creation is not a free template."

These strong words were quoted by several media outlets, including Reuters.BWorld Online

Furthermore, Holzmüller stated,

"Today, we have set a precedent to protect and clarify the rights of authors. Even operators of AI tools like ChatGPT must comply with copyright law."

He emphasized that this trial is a victory for protecting the livelihoods of creators.The Guardian

TechCrunch also highlighted this comment, portraying it as a "battle to protect the livelihoods of music creators."TechCrunch

3. What Was Illegal? A Rough Breakdown of the Legal Points

Legally, the issues can be broadly categorized into three points.

1. Unauthorized use of lyrics as training data
   Both German copyright law and the EU Copyright Directive have mechanisms that allow "text and data mining (TDM)," but legal experts often interpret this as limited to cases where "rights holders have not opted out" and for "temporary analysis purposes."Lexology

2. The model "almost verbatim" memorized the lyrics
   Ideally, a trained model should abstract "patterns" to generate new text, but in practice, a phenomenon known as memorization occurs, where parts of the data are "remembered" almost verbatim. GEMA demonstrated that ChatGPT could reproduce lyrics quite faithfully, arguing that "this is not mere learning but the preservation and re-offering of the work."Reuters

3. It was judged not to fall under TDM "exceptions"
   According to the ruling's explanation, the court determined that "training an AI model is not mere information extraction but an act of permanently embedding reproductions of works within the model, exceeding the scope of TDM exceptions."Lexology

In summary,

"Massively reading in lyrics without permission, storing them almost entirely within the model, and making them available for redistribution to users is not an exception to copyright."

This is the line that has been drawn.

4. Reactions on Social Media: "The First Day Supporting GEMA" and Concerns About AI

As soon as this news broke, various reactions swirled on social media platforms like X (formerly Twitter), Reddit, and Instagram.

4-1. Surprise at "Supporting GEMA for the First Time"

In Germany, GEMA is often criticized by users for reasons such as "collecting fees even for club events" and "blocking content on YouTube."

However, in the German-speaking Reddit community discussing this ruling,

comments like "For the first time in my life, I'm siding with GEMA"

were gathering likes.Reddit

On the other hand, within the same thread,

there were also sharp comments like "Isn't GEMA itself an organization that profits by hanging on to human creations?"

indicating a sense of distrust not just towards AI vs. creators, but towards the entire rights business.Reddit

4-2. Legal Experts Focus on the Recognition of Both "Memory and Output" as Infringement

On X, accounts knowledgeable about tech law and copyright summarized this ruling as follows:

"The German court recognized both the internal memorization within the model and the chatbot output as infringements. A turning point for AI companies."

This post was shared, sparking discussions among overseas legal experts and AI researchers.X (formerly Twitter)

What is being focused on here is the extent to which the state of "embedding works in the model's weights" itself is considered problematic from a copyright perspective. There is analysis suggesting that judgments are diverging between Germany and the UK, and how these differences in national stances will affect open-source AI and multi-region deployment is becoming a focal point.THE DECODER

4-3. General User Voices: "Putting the Brakes on AI" and Concerns About "AI Fleeing Europe"

On Instagram and in the comment sections of news media,

• comments like "This finally sets a precedent that AI companies can't just 'free ride'"

• and "It's only natural to protect artists' work"

were prominent, showing support for creators.

• On the other hand, concerns like "What if AI companies withdraw from Europe all at once because of this?"

• and "Won't this put a damper on innovation?"

were also strongly expressed.Instagram

Overall, a clear conflict structure is emerging between those welcoming "rights holder protection" and "strengthening AI regulation" and those fearing that "excessive regulation will strangle startups and researchers."

5. How Does the German Ruling Differ from Global AI & Copyright Lawsuits?

OpenAI is facing multiple copyright lawsuits, not just the GEMA lawsuit, primarily in the United States. Notable cases include

• a lawsuit over unauthorized learning of news articles by The New York TimesReuters

• a class action by authors' groups like the Authors GuildThe Authors Guild

• a 2025 lawsuit by Ziff Davis, a media company owning PCMag and IGNReuters

Furthermore, disputes between AI companies and the content industry are becoming a global trend, with companies like Anthropic being sued by music publishers over lyrics usage.The Guardian

Many of these U.S. lawsuits focus on the abstract and comprehensive debate of whether **"AI learning constitutes fair use."**
In contrast, the German ruling

• focused on the specific lyrics of certain songs (nine songs)

• their use in learning

• and their actual output in "almost the same form