ChatGPT came up with a 'Game of Thrones' sequel idea. Now, a judge is letting George RR Martin sue for copyright infringement. (www.businessinsider.com)

A federal judge in Manhattan has allowed a consolidated class-action suit from authors — including George R.R. Martin, Michael Chabon, Ta-Nehisi Coates, Jia Tolentino and Sarah Silverman — to proceed against OpenAI and Microsoft, finding that a ChatGPT-generated outline for a proposed sequel to Martin’s A Song of Ice and Fire could be “substantially similar” to the author’s work. The ruling cites a sample prompt asking ChatGPT to create “an alternative sequel to ‘A Clash of Kings’” and the model’s response, which proposed a book called A Dance with Shadows and suggested plot elements such as a novel “ancient dragon-related magic,” a distant Targaryen heir named Lady Elara, and a rogue sect of the Children of the Forest. Judge Sidney Stein said a reasonable jury could find copyright infringement based on such outputs. The decision matters because it moves a high-profile technical-legal question closer to trial: whether LLM makers can ingest copyrighted books without permission and whether their generated outputs infringe those works, or are protected as fair use. The judge did not resolve the fair-use defense — an unresolved issue after a San Francisco judge found Anthropic’s training was fair use (Anthropic later settled for $1.5B). Practically, the case could shape how models are trained, compel licensing or stricter filtering of training corpora, and influence how companies handle prompts and content controls if courts impose broader liability for model outputs.