Copyright Winter Is Coming (To Wikipedia?) (authorsalliance.substack.com)

A federal judge in Authors Guild v. OpenAI (S.D.N.Y.) denied OpenAI’s motion to dismiss claims that ChatGPT outputs infringe the rights of authors including George R.R. Martin and David Baldacci, finding that a ~580‑word ChatGPT summary of A Game of Thrones could plausibly be “substantially similar” to the novel. The court accepted the plaintiffs’ submitted outputs as part of the complaint and concluded that the summary “parrots” plot, characters and themes and “attempts at abridgment or condensation” of central copyrightable elements. The ruling is noteworthy because it treats short, analytic summaries as potentially infringing at the pleading stage even though the judge expressly left fair use and final liability for later resolution. For AI/ML developers, publishers and platforms, the decision is a red flag: it narrows the idea–expression distinction as applied to fiction and suggests that concise model-generated summaries — and by extension comparable Wikipedia plot entries or other condensed analyses — could be swept into infringement disputes. Key technical points include the court’s reliance on output-level comparison (not just training process), the similarity between the ChatGPT summary and an ~800‑word Wikipedia plot (despite structural differences), and the fact that the ruling concerns plausibility, not ultimate guilt. The practical implication is increased legal risk for generative models that reproduce narrative structure, characters, or plot arcs, and a potential chill on automated summarization, knowledge bases, and derivative-content tooling unless fair use or higher‑court clarifications intervene.