Both sides seek a ruling without trial in AI book case

The Authors Guild, its co-plaintiff authors, OpenAI and Microsoft have all asked Judge Sidney Stein to decide the book-training copyright case without a trial. OpenAI filed its summary judgment motion on September 4 and the authors filed theirs on September 5, Publishers Weekly reports. Both sides’ filings turn on fair use.

A summary judgment motion asks the court to rule as a matter of law, even if it accepts the other side’s version of the facts as true. If the court finds that material facts are still in dispute, the case goes to trial, as the Authors Guild explains in its post on the filing. The authors ask Stein to find that OpenAI and Microsoft infringed in three ways: by pirating the works, by copying them to train AI, and by trading the books between themselves as currency. They also ask him to reject the fair use defence outright.

“Mass piracy”

The plaintiffs’ brief opens by calling OpenAI’s GPT models “an existential threat to those who write and publish books”, one that “threatens the incentives to create, the livelihoods of authors, and the future of book publishing.” Its central charge is that “OpenAI built the foundations of its business on mass piracy”. That includes book content reproduced from the pirate site Library Genesis (LibGen) and later deleted.

The deletion became its own fight in discovery. In 2022 OpenAI deleted two training datasets, books1 and books2, which are believed to come from LibGen and other shadow libraries. The plaintiffs argue the deletion looks designed to hide what the models were trained on, and in late 2025 a magistrate judge ordered OpenAI to produce its internal Slack messages about it. On acquisition, the authors’ brief says courts have “consistently held that obtaining ‘for free’ something one would ‘ordinarily have to buy’ is not fair use.”

Much of the brief is blacked out. The redactions cover the number of books affected and the methods by which, according to the plaintiffs, OpenAI and Microsoft “worked together” to copy copyrighted material and build training datasets from it. One passage says books are “high-quality data because they contain organized writing, structured thoughts, and” a further list of characteristics, which is redacted.

On harm, the Guild says the brief calls the effect on the market the “single most important element” of fair use and argues that “OpenAI’s models are capable of displacing human authors precisely because it exploited their works.” As evidence, the plaintiffs point to rising sales of AI-generated fiction on platforms such as Kindle. They also cite Tarun Gogineni, whom the brief identifies as OpenAI’s Head of Model Writing Quality and whose “research mission”, it says, was to have GPT models write the “last two books” of George R.R. Martin’s A Song of Ice and Fire. He is also quoted from a viral 2025 tweet: if Martin “dies early, GPT-5 will autocomplete his series.”

0.62% of A Game of Thrones

OpenAI’s September 4 filing argues the use was fair “because it was highly transformative”, and that “no property right is unlimited.” Books ended up in the training data, the defendants say, because “training a useful LLM requires hundreds of billions to tens of trillions of words from a multitude of diverse sources”. They say the text is mined for general patterns of language, not to reproduce “creative expression”.

On output, the defendants say ChatGPT “does not display copies” of books and put its “alleged regurgitation rate” at 0.00007%. In their account, word-for-word sequences outside idioms or public-domain texts such as the Gettysburg Address are rare. The plaintiffs’ expert, they write, “spent significant effort trying to brute force OpenAI’s LLMs to reproduce text” with millions of prompts. “The longest span of contiguous text he could generate was a 1,899-word excerpt from A Game of Thrones, which is only 0.62% of the book.”

On market harm, OpenAI argues that future books may be written with AI help or without it, and that “competitive harm from new, non-infringing works is not harm that copyright protects against.” The deposition evidence in this section is mostly redacted in the public version. The defendants compare their position to Authors Guild v. Google Inc. (Google Books) in 2015 and Kadrey v. Meta Platforms in 2025, where the defendants won on fair use. The plaintiffs argue that, unlike in those two cases, this copying was “nontransformative”. In their account it was done in secret “for the purpose of developing a competing commercial product”, and it took away the chance for authors and publishers to earn money from AI training licences.

How the case got here

The docket, now called In Re: OpenAI Inc. Copyright Infringement Litigation (1:25-md-03143), is before Stein in the Southern District of New York. The book claims are being heard alongside the news organizations’ case led by the New York Times.

• September 19, 2023: the Authors Guild and 17 authors sue OpenAI in the Southern District of New York.
• December 2023: Microsoft is added as a defendant.
• April 3, 2025: the case is consolidated with other pending class actions and centralized in a multi-district litigation with the New York Times’ and other news companies’ suits.
• October 27, 2025: Stein denies OpenAI’s motion to dismiss the output-infringement claims. He finds a reasonable jury could find ChatGPT’s summaries of the plaintiffs’ novels substantially similar to the books, and does not rule on fair use.
• February 27, 2026: fact discovery closes. The Guild says expert discovery was finished in summer 2026.
• September 4-5, 2026: both sides file cross-motions for summary judgment.

Thirteen authors are now lead co-plaintiffs: David Baldacci, Taylor Branch, Michael Connelly, Sylvia Day, Jonathan Franzen, Christopher Golden, Andrew Sean Greer, John Grisham, David Henry Hwang, George R.R. Martin, Jodi Picoult, Stacy Schiff and James Shapiro. Baldacci, Connelly, Day, Franzen, Grisham, Martin and Picoult have been in the case since it was filed.

The October 2025 ruling dealt with what ChatGPT produces and left fair use undecided; both sides’ motions now ask Stein to rule on it. Oppositions were due on October 9 and replies are due on November 6. No hearing date has been set.