CONTENTS
- 1. Copyright Infringement Litigation | The Problem of Unauthorized Use of AI Training Data
- 2. Copyright Infringement Litigation | Analysis of the OpenAI and Anthropic Cases
- - Germany, the Judgment Finding OpenAI's ChatGPT Liable for Copyright Infringement
- - The United States, the Large-Scale Settlement in the Class Action Against Anthropic
- 3. Copyright Infringement Litigation | An In-Depth Analysis of the Courts' Reasoning
- - The U.S. Court's Application of the ‘Fair Use' Doctrine
- 4. Copyright Infringement Litigation | Implications for AI-Related Companies
1. Copyright Infringement Litigation | The Problem of Unauthorized Use of AI Training Data
Copyright infringement lawsuits are occurring frequently around the world.
As artificial intelligence technology has recently advanced at an unprecedented pace, the problem of the unauthorized use of copyright-protected content contained in the vast data that AI uses for training is emerging as a significant legal issue worldwide.
In particular, the AI copyright infringement cases that have arisen in the United States and Germany are having a decisive influence on setting the standards for the future ecosystem of the AI industry and for the protection of creators' rights.
2. Copyright Infringement Litigation | Analysis of the OpenAI and Anthropic Cases
This article examines AI copyright infringement litigation, focusing on the judgment of the Munich Regional Court in Germany concerning OpenAI's ‘ChatGPT' and the class-action settlement involving Anthropic before the U.S. federal court in California.
Germany, the Judgment Finding OpenAI's ChatGPT Liable for Copyright Infringement
On November 11, 2025, the Munich Regional Court in Germany held that OpenAI's ‘ChatGPT' had violated copyright law by using, without authorization, the lyrics of nine songs, including a song by the renowned German musician Herbert Grönemeyer, as training data.
This copyright infringement lawsuit was filed by the German music copyright society (GEMA) against OpenAI.
GEMA, which represents the rights of some 80,000 member creators, including composers and lyricists, challenged the unlawfulness of the unauthorized use of the data.
The court assessed, among other things, whether the act of training the AI model on the data itself infringed the copyright holders' reproduction rights and rights of use, and found that ‘OpenAI's act of storing the copyright-protected lyrics and reproducing them verbatim when needed' constituted clear copyright infringement.
- OpenAI's rebuttal : It argued that the AI does not store specific data but only learns patterns from a wide range of data, so responsibility should rest with the user.
- The court : The act of memorization at the AI training stage itself, and the very possibility of reproducing the work, serve as grounds for copyright infringement.
Accordingly, the court ordered OpenAI to cease storing the lyrics and outputting them in its responses, together with an award of damages.
OpenAI, however, is expected to appeal the decision.
The United States, the Large-Scale Settlement in the Class Action Against Anthropic
In the United States, the AI company Anthropic became embroiled in a large-scale copyright infringement lawsuit with authors over allegations that it had downloaded numerous pirated books in bulk and used them to train its own AI model, ‘Claude.'
In response, the U.S. District Court for the Northern District of California granted preliminary approval to a settlement worth several billion dollars.
The court recognized that the doctrine of fair use under U.S. copyright law could apply to training that relied on books Anthropic had lawfully purchased.
As to the portion involving the unauthorized use of pirated books, however, it found copyright infringement and approved the large-scale settlement.
Anthropic is reported to have submitted a settlement proposal offering to pay approximately 3,000 dollars per book plus interest and to destroy the data set at issue.
3. Copyright Infringement Litigation | An In-Depth Analysis of the Courts' Reasoning
The German court's judgment reflects Europe's stringent stance on copyright protection in the face of advancing AI technology.
The court characterized the very act of ‘memorizing' copyright-protected lyrics during the AI training process and ‘reproducing' or ‘outputting them verbatim' when needed as a reproduction that infringes the copyright holders' rights of use.
This appears to be a legal interpretation under which, where an AI model does not merely extract patterns from data but in effect has the capacity to store and reproduce the core portions of the original work, the act of training itself can be regarded as copyright infringement.
This judgment is consistent with the European Union (EU) copyright directive, and it is expected to bring about changes in how AI companies in Europe select and manage their training data going forward.
The U.S. Court's Application of the ‘Fair Use' Doctrine
In the Anthropic case, the U.S. court made its determination by strictly applying the ‘fair use' factors set out in Section 107 of the U.S. Copyright Act.
Accordingly, it did not find the act of using copyright-protected works for LLM training, in itself, to be unlawful.
As for Anthropic's direct purchase and scanning of printed copies, the court recognized fair use, because the copying was done to create digital copies that save storage space and make the works searchable.
However, it found copyright infringement with respect to the act of downloading and retaining ‘pirated copies' of the books, ruling in favor of the authors and granting preliminary approval to the settlement.
Section 107 of the U.S. Copyright Act
The United States provides for the following four fair-use factors as exceptions to copyright infringement.
① the purpose and character of the use
② the nature of the work used
③ the amount and substantiality of the portion used
④ the effect of the use upon the potential market for or value of the work used
Thus, it provides that, where the use is noncommercial or for an educational purpose and the work used has no adverse effect on its potential market, or where the portion used does not exceed the amount of the original work, among other requirements, using a work without the copyright holder's permission does not constitute copyright infringement.
4. Copyright Infringement Litigation | Implications for AI-Related Companies
The major judgments and settlement cases in the two countries are likely to serve as a starting point for establishing legal implications and practical response strategies for every company that develops or uses AI technology.
1) Securing the ‘Legality' of Data Acquisition
2) Managing Copyright in AI Training and Outputs
3) Strategic Response to Global Regulatory Trends
The Munich Regional Court's judgment concerning OpenAI and the U.S. federal court in California's settlement involving Anthropic show that, for the AI industry, the seriousness of copyright issues and the scope of legal responsibility are no longer areas that can be overlooked.
In a situation where securing the legal legitimacy of the data that forms the basis of AI technology has become an indispensable precondition for a company's sustainable growth, you are advised to remain aware of global copyright infringement litigation trends and to make active use of legal counsel on copyright-related matters.
If you need a more detailed assessment of legal risks or tailored advice in connection with operating an AI business, you are welcome to contact us at any time.
