Penguin Random House Verlagsgruppe has launched legal proceedings in Munich against OpenAI’s European arm, alleging that the company’s chatbot reproduced and imitated the copyrighted works of German author-illustrator Ingo Siegner. The publisher says that, in response to a targeted prompt asking the model to write a children’s book placing Siegner’s character Coconut the Little Dragon on Mars, the system produced both text and images that closely matched the tone, art and packaging of the established series.

According to the publisher, the output included a narrative, a front cover showing an orange dragon accompanied by two sidekicks, a back-cover blurb and even step-by-step guidance on submitting the manuscript to a self-publishing platform. Penguin Random House argues these results demonstrate a form of “memorisation” by the large-language model, in which substantial portions of training material are effectively reproduced rather than merely informing general patterns.

The case follows recent German court scrutiny of generative AI. Last November a Munich court found in favour of the country’s music rights society, concluding that ChatGPT had used protected song lyrics in ways that violated copyright law. Legal observers say the new lawsuit, brought by one of the world’s largest publishers, could sharpen the legal framework for how training data and model outputs are treated across creative industries.

Penguin Random House’s children’s and young-adult publishing head, Carina Mathern, said the company supports exploring AI’s possibilities but must defend authors’ rights and livelihoods. An OpenAI spokesperson responded that the company was reviewing the complaint and that it engages in discussions with publishers globally to seek ways for creators to benefit from AI technology. The publisher’s complaint was filed at the Munich regional court and notes that earlier commercial arrangements between OpenAI and media groups did not grant access to certain archival content.

The dispute raises practical and technical questions about how generative models are trained and how close an AI-generated work must be to a pre-existing piece to amount to infringement. Industry commentary highlights two faultlines: whether replication represents unlawful copying or lawful statistical synthesis, and whether licensing, opt-out mechanisms or technical limits on memorisation can reconcile innovation with copyright protection.

As the litigation progresses, its outcome is likely to be watched by publishers, creators and AI firms worldwide as courts test how existing intellectual property law applies to generative systems. Agence France-Presse and multiple German media outlets have reported on the filing and its potential implications for authors and illustrators whose work may have been used in model training.

Source Reference Map

Inspired by headline at: [1]

Sources by paragraph:

Source: Noah Wire Services