Is ChatGPT allowed under EU copyright law to generate a 'Game of Thrones' sequel?

Sadly, there won't be a ninth season of "Game of Thrones" anytime soon. However, in the US, ChatGPT has written a sequel to one of the underlying books, and a US court is hearing the case on grounds of copyright infringement. How would a European court assess such a case?

“Game of Thrones” is one of the most famous TV series in the world. So far, there have been eight seasons and many spin-offs. There is speculation as to whether there will be a ninth season. The author of the books on which the series is based, George R. R. Martin, has not yet made a final decision on this. So why not write the ninth season yourself? ChatGPT could help. But would that be permitted by copyright law?

As the following case demonstrates, this is a far from unreasonable question. The District Court of the Southern District of New York, a federal court of first instance in the US, is currently dealing with a sequel to 'A Clash of Kings', the second volume in George R. R. Martin's book series 'A Song of Ice and Fire'.. This book forms the basis for the second season of Game of Thrones. Martin's lawyers asked ChatGPT to provide a detailed outline of a sequel to “A Clash of Kings”. The result was a story called 'A Dance with Shadows'. ChatGPT incorporated several of Martin's fictional key characters, such as Tyrion, Sansa, Robb, Daenerys, Cersei and Jon, and developed plots that continued the storylines of 'A Clash of Kings', introducing new factions, magic and claims to the Iron Throne. George R. R. Martin is claiming copyright infringement, while OpenAI is arguing that it is a permitted use under US copyright law, i.e. fair use.

This case seems similar to the GEMA v OpenAI lawsuit, which was ruled on by the Munich I Regional Court in November 2025. In that case, OpenAI, the operator of ChatGPT, was found to have infringed copyright. ChatGPT repeatedly produced various lyrics of well-known German songs almost identically through simple prompts, such as 'Atemlos' by Helene Fischer (court ref. 42 O 14139/24). According to the court, this constituted an unlawful reproduction and public disclosure of the lyrics. But what about using ChatGPT to write serial stories and the implications this has under European copyright law?

Novel sequels are usually subject to licensing requirements

Under EU copyright law sequels or other spin-offs of fictional material are subject to the consent of the copyright holders. The German Federal Court of Justice (BGH) ruled on this for German law more than 25 years ago in its famous “Lara's Daughter” decision (judgment of April 29, 1999, court ref. I ZR 65/96). In this case, the German BGH found copyright infringement because the novel “Lara’s Daughter” (German: “Laras Tochter”) had continued Boris Pasternak’s novel “Doctor. Zhivago” without the consent of “Dr. Shiwago’s” publisher.

The considerations of the German BGH are still relevant for EU law today. In its recent Mio and konektra ruling, the CJEU held that an adaptation occurs whenever a protected element of a work remains recognizable within another (judgment of December 4, 2025, ref. C-580/23, C-795/23 – Mio and konektra). Furthermore, this decision confirmed that an adaptation, as a rule, inherently involves an act of reproduction. As soon as an element protected by copyright is recognisable in the adaptation, this constitutes copyright infringement unless a licence or other consent has been granted.

Furthermore, unlike texts with real content, fictional texts are also protected in terms of their content and not only in terms of their form. Therefore, copyright protection does not end with the external form of the text; it also covers the content, provided it reflects the creator's personality. If such content is continued, copyright-protected elements are adopted. The continuation of the stories is dependent on the incorporation of fictional content. If you apply CJEU Mio and konektra, this means that sequels constitute an adaptation (reproduction) that requires consent.

It could be a permissible 'pastiche', but what exactly is that?

It is conceivable that the continuation of a novel could be permitted under EU copyright exceptions. The limitation for caricature, parody and pastiche (Article 3, paragraph 3(k) Directive 2001/29) is particularly relevant here. It cannot be ruled out from the outset that a sequel could be a permissible parody. For this to be the case, the sequel must evoke the original work while being noticeably different. Further, it needs to contain a humorous element. Furthermore, the CJEU stipulates that a comprehensive balancing of interests must be carried out between the interests of copyright holders and the protection of freedom of expression (CJEU judgment of September 3, 2014, ref. C-201/13 – Deckmyn/Vrijheidsfonds/Vandersteen). According to the jurisprudence of the German Federal Court of Justice (BGH), the balancing of interests will likely favour the continuation of the parody, especially if it is directed against the original (judgment of July 28, 2016, ref. I ZR 9/15 – Auf Fett getrimmt).

That leaves 'pastiche'. The meaning of 'pastiche' will be clarified in a preliminary ruling procedure before the CJEU, initiated by the First Civil Senate of the German Federal Court of Justice, which is responsible for copyright law ('Pelham', C-590/23, opinion by the Advocate General Emiliou available, CJEU judgment expected for 14 April 2026). The comparative use often required for a permissible pastiche could perhaps be provided in a sequel. That said, there are serious doubts as to whether the so-called three-step test excludes a permissible pastiche in the continuation of fictional texts. The three-step test must be considered whenever a limitation is applied (Article 5(5) of the InfoSoc Directive 2001/29). This ensures that exceptions granted by copyright protection are proportionate, and it will play a significant role in limiting the scope of pastiche. Applying to the three-step test, this should rule out a legal pastiche. A sequel interferes with the normal exploitation of the work. This is because normal exploitation of a work involves the author controlling the continuation. Examples include the Game of Thrones books, the seven Harry Potter novels, and the Asterix comics.

Who is the infringer?

Overall, sequels to fictional novels and other works of fiction are generally considered copyright infringement under EU copyright law if they are created without the consent of the rights holders. But does the fact that a generative AI such as ChatGPT writes such a sequel change anything? Most likely, it does not.

Creating the sequel is still a copyright infringement. However, this raises the question: Who is liable as the (primary) infringer? As is often the case in law, a distinction must be made – it depends! The party that primarily determines the content is liable as the infringing party. If the AI is determining the content, for example because the user has provided no specifications in the form of specific and detailed prompts, then the AI operator should be the infringer. In the GEMA/Open AI case, the Munich I Regional Court ruled that OpenAI was liable as the reproducer because the song lyrics in question could be repeatedly retrieved from ChatGPT using easy and simple prompts (court ref. 42 O 14139/24). The same should apply to the continuation of fictional texts written by generative AI.

Liability for publishing (distribution and/or making available to the public) of the sequel is assessed differently under EU copyright law. Such public use of the AI output constitutes a copyright infringement and is attributable to the AI users who distribute or make it available to the public. Such infringement occurs, for example, if the AI users make the AI-generated sequel novel publicly available on the internet or have it printed and distributed as a book.

Current copyright law is shown to be capable of dealing with the challenges of the AI age through the example of fictional texts being continued by generative AI such as ChatGPT. Such continuations are generally under the legal control of the copyright holders. You need the consent of the copyright holder to use them. But there are certain exceptions — for example, in the case of certain parodies. Following the decision in the GEMA/OpenAI case in Germany, it can be assumed that OpenAI itself infringes copyright (primary liability) if it produces fictional continuations of copyrighted material in a repeatable manner in response to simple prompting.

This article is an English adaptation of the German article 'Durfte ChatGPT eine Fortsetzung von Game of Thrones schreiben?', written by the author and originally published on Legal Tribune Online.