In a landmark ruling on 11 November 2025 (case no. 42 0 14139\/24), the Munich Regional Court found OpenAI, in its capacity as an AI model operator, liable for copyright infringement. The case centred on the \u201cmemorisation\u201d of song lyrics by large language models (LLMs) and their near-verbatim reproduction in response to basic user prompts.<\/p>\n\n\n\n
Both the reproduction (\u201cmemorisation\u201d) of the song lyrics by the language models and their communication in chatbot output constitute infringements of copyright exploitation rights. According to the court, these are not covered by any legal limitations, particularly those related to text and data mining (TDM), which are restricted to preparatory copies, provided these are strictly necessary for analytical purposes.<\/p>\n\n\n\n
At its core, the dispute centres on allegations that OpenAI\u2019s GPT\u20114\/4o models reproduced substantial portions of the lyrics of nine well\u2011known German songs in response to basic prompts, indicating that they had memorised the training data. The Munich court\u2019s decision is the first German ruling to address the extent to which the training and output of LLMs may have copyright implications. The court took a very narrow view, rejecting the application of copyright law exemptions for LLM operators. In cases where basic prompts produce near-verbatim or identical reproductions of copyrighted content, the court held model operators, rather than users, responsible.<\/p>\n\n\n\n
As an interesting contrast, just days earlier, on 4 November 2025, the London High Court came to the opposite conclusion on a key point in Getty Images v Stability AI, dismissing Getty\u2019s secondary infringement claim on the basis that the Stable Diffusion model did not constitute an \u201cinfringing copy\u201d. This reflects a fundamentally different view of LLM technicalities and their copyright implications.<\/p>\n\n\n\n
While the Munich court\u2019s ruling is significant, it merely marks the beginning of the legal clarification process. The judgment is not yet final, and the proceedings are far from over. The core issues \u2013 what counts as \u201creproduction\u201d by an LLM, the scope of TDM exemptions and operator liability for outputs \u2013 remain unresolved. Appeals can be expected all the way up to the German Federal Court of Justice (Bundesgerichtshof) and potentially even the Court of Justice of the European Union.<\/p>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
In a landmark ruling on 11 November 2025 (case no. 42 0 14139\/24), the Munich Regional Court found OpenAI, in its capacity as an AI model operator, liable for copyright infringement. The case centred on the \u201cmemorisation\u201d of song lyrics by large language models (LLMs) and their near-verbatim reproduction in…<\/p>\n","protected":false},"author":49,"featured_media":56531,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"inline_featured_image":false,"footnotes":""},"categories":[959],"tags":[991,992],"ausgabe":[669,675],"verknuepfung":[935,718,755,766],"class_list":["post-56529","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-short-news","tag-ai","tag-artificial-intelligence","ausgabe-gsk","verknuepfung-digitalisation","verknuepfung-dispute-resolution","verknuepfung-data-protection","verknuepfung-ip-it-en"],"acf":{"ansprechpartner":[{"anwalt":891}],"standard_text":true,"custom_text":false,"postdownloads":false,"pressekontakt":{"email":"","phone":""}},"info":{"date":"25.11.2025","author":"Andrea Gudera","categories_names":["Short News"],"category":"Short News"},"wpml_main_post_id":56528,"featured_img_src":"https:\/\/gsk.de\/wp-content\/uploads\/2025\/11\/GSK-Stockmann-Shortnews-KI-Training.jpg","_links":{"self":[{"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/posts\/56529","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/users\/49"}],"replies":[{"embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/comments?post=56529"}],"version-history":[{"count":2,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/posts\/56529\/revisions"}],"predecessor-version":[{"id":56552,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/posts\/56529\/revisions\/56552"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/media\/56531"}],"wp:attachment":[{"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/media?parent=56529"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/categories?post=56529"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/tags?post=56529"},{"taxonomy":"ausgabe","embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/ausgabe?post=56529"},{"taxonomy":"verknuepfung","embeddable":true,"href":"https:\/\/gsk.de\/en\/wp-json\/wp\/v2\/verknuepfung?post=56529"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}