A Munich Regional Court filing puts two distinct legal theories about AI copyright in front of a German judge for the first time at this scale. According to the Mishcon de Reya AI litigation tracker, Penguin Random House initiated proceedings against OpenAI on approximately March 27, 2026, centered on children’s author Ingo Siegner’s work and ChatGPT’s alleged reproduction of content from that catalog.
The tracker characterizes the claim as addressing two separate issues: the use of copyrighted material during training, and the generation of outputs that reproduce or closely resemble the protected content. These aren’t the same legal theory. The first is a question about whether training on copyrighted works without authorization violates copyright law, the core issue in dozens of ongoing AI copyright cases globally. The second is distinct: whether a model that generates content resembling a specific author’s work infringes copyright independently of how the model was trained. A court could find on one theory and not the other.
That structure is what makes this case worth watching beyond its specific facts. If the Munich Regional Court accepts both theories, it would establish a German precedent that AI developers could face liability at training time and at inference time, two separate exposure points requiring two separate compliance responses. That’s a materially different risk landscape than US copyright cases, which have largely focused on the training data question under fair use analysis.
German copyright law doesn’t offer a direct equivalent to US fair use doctrine. The applicable framework in Germany involves specific statutory exceptions, and the question of whether AI training falls within them has not been settled by a German court at this level. The Munich proceedings will generate analysis, and possibly a ruling, that informs how EU jurisdictions approach AI copyright before any EU-level mandatory framework is in place.
This case joins a field of active copyright proceedings against AI companies across multiple jurisdictions. The hub has covered the EU legislative dimension of this issue, EU Parliament’s call for mandatory AI copyright rules and the four-stakeholder Brussels copyright debate, but litigation is moving faster than legislation in several jurisdictions.
What to watch: the Munich Regional Court’s initial procedural decisions will indicate whether the case proceeds on both theories or is narrowed early. Watch also for OpenAI’s jurisdictional arguments, whether a US company can be compelled to answer for training and inference practices under German law is itself a contested question. The case update tracked by Mishcon de Reya as of April 27 suggests the proceedings are progressing, though the specific procedural milestone isn’t clear from available tracker information.
All characterizations of this case come from the Mishcon de Reya AI litigation tracker, a single T2 source. Court documents haven’t been independently confirmed. Treat the dual-theory framing as tracker analysis, not court-verified fact.