Escape Into Books
Your home for book news, reviews, and bookish fun. Follow along on Facebook.
A federal judge says authors can press ahead with a major copyright lawsuit against OpenAI and Microsoft, citing a ChatGPT-generated sequel idea for George R.R. Martin’s fantasy epic as a key example of potential infringement.
In a ruling issued Monday in Manhattan, US District Judge Sidney Stein said a reasonable jury could find that some ChatGPT outputs are “substantially similar” to the novels they resemble, including Martin’s still-unfinished A Song of Ice and Fire series. The decision keeps alive a consolidated class action brought by authors including Martin, Michael Chabon, Ta-Nehisi Coates, Jia Tolentino, and Sarah Silverman.
ChatGPT’s ‘Game of Thrones’ sequel under the microscope
The case turned, in part, on a prompt asking ChatGPT to outline an alternative sequel to Martin’s A Clash of Kings that diverged from the published third book, A Storm of Swords. According to court records reported by Business Insider, the chatbot enthusiastically proposed a new volume titled “A Dance with Shadows,” complete with fresh battles for the Iron Throne, a distant Targaryen relative named Lady Elara, and ancient dragon-linked magic.
For Stein, those details were enough to let the authors’ core claim move forward: that the AI company ingested their books without permission and produced outputs that echo their protected worlds and characters. The ruling follows an earlier Authors Guild lawsuit that first put Martin and fellow bestseller John Grisham on a collision course with OpenAI in 2023, when they alleged the company trained its models on their books without consent or payment, as outlined by CNBC.
It also arrives as Martin’s work faces AI pressures on other fronts, from spinoff TV projects to debates over fan creations. Recent coverage of new Game of Thrones spinoff ideas and his ongoing battle to finish The Winds of Winter shows just how closely fans track every twist in his universe—whether it’s crafted by the author himself or generated by a chatbot.
Fair use questions and a fractured legal landscape
The New York case is part of a wider legal reckoning over how AI companies train massive language models. Earlier this year, multiple author and newspaper suits against OpenAI and Microsoft were merged in New York because they raise the same core question: can tech firms scrape books, articles, and other texts at scale without licenses? The Guardian reports that these combined cases could help set nationwide ground rules.
So far, courts have split. A San Francisco judge previously found that Anthropic’s use of copyrighted books to train its Claude model was protected fair use, though the company later agreed to a massive $1.5 billion settlement with authors. In June, Meta scored a win when a judge sided with its fair use defense in an AI training lawsuit brought by writers, as detailed by The Guardian.
Judge Stein hasn’t yet ruled on whether OpenAI and Microsoft can claim similar fair use protection, or whether the authors will be certified as a class. Those decisions will come later. For now, the ruling means discovery can move ahead—and with it, more scrutiny of exactly how ChatGPT and its successors learned to write so convincingly like our favorite storytellers.
The outline the judge focused on went further than a change of title. Alongside Lady Elara and the dragon-linked magic, ChatGPT’s “A Dance with Shadows” invented a rogue sect of the Children of the Forest, dropping whole new factions into Martin’s existing map. The Authors Guild, which organised the class actions behind the case, alleges the models were trained on full-text novels taken from pirate ebook repositories rather than licensed copies.
Pressure on OpenAI is building elsewhere in the same courthouse. In December 2025 a judge ordered the company to hand over internal ChatGPT logs to The New York Times, which accuses it of copying and regurgitating its journalism, Reuters reported. Researchers have separately shown that models fine-tuned on an author’s full catalogue can turn out convincingly on-brand new work in that voice, which is precisely the outcome the plaintiffs say they never agreed to.
For readers, the stakes are more than abstract. These cases will help decide whether beloved novels can be quietly fed into AI systems without permission, and who—if anyone—gets paid when a chatbot dreams up the next would-be epic in the style of Westeros.
