Penguin Random House filed a lawsuit in Munich last week against OpenAI, alleging that ChatGPT reproduced the content of "The Little Dragon Coconut" (Der kleine Drache Kokosnuss), a popular German children's book series, in a manner the publisher describes as "virtually indistinguishable" from the originals.

According to the lawsuit, when prompted to generate content from the series, ChatGPT produced not just matching text but also a book cover and back-cover blurb that the publisher says copied the originals without authorization. The case was filed in Munich Regional Court, placing it under German copyright law, which provides strong protections for creative works regardless of the medium of reproduction.

The "virtually indistinguishable" standard is significant. Most AI copyright suits have argued that training data use is the violation. Penguin's Munich case argues something different: that the output itself reproduces protected expression, which is a cleaner legal theory and potentially a more dangerous one for OpenAI. If ChatGPT can reproduce a children's book nearly in full when prompted, the question of whether training on that book constitutes fair use becomes secondary to whether the reproduction of it does.

OpenAI has not issued a substantive public response to the Munich filing.

-- ANNA WEBER, Berlin

Sources: