George R. R. Martin’s AI Lawsuit Could Change How Books Train AI

George R. R. Martin and a group of prominent authors have reached a potentially decisive stage in their copyright lawsuit against OpenAI and Microsoft.

On September 4, 2026, the authors asked U.S. District Judge Sidney H. Stein to rule that the companies infringed their copyrights by allegedly obtaining their books without permission and copying them to train artificial-intelligence models. OpenAI and Microsoft filed competing motions asking the judge to accept their fair-use defense.

The dueling requests for summary judgment move the litigation closer to an answer on one of the biggest unresolved questions facing the publishing industry: Can an AI company copy copyrighted books to train a commercial model without first obtaining permission from, or paying, the authors?

No ruling has been issued, and the allegations against OpenAI and Microsoft have not been proven. Nevertheless, the judge’s eventual decision could influence how authors, publishers and technology companies negotiate the future of AI training.

What Just Happened in the Lawsuit?

The Authors Guild and its co-plaintiffs filed a motion for partial summary judgment on September 4. The filing asks Judge Stein to decide important parts of the case based on the developed factual record rather than sending every question to a trial.

The authors’ case is now part of the consolidated litigation known as In re OpenAI, Inc., Copyright Infringement Litigation, pending in the Southern District of New York.

Along with Martin, the named authors include David Baldacci, Michael Connelly, Jonathan Franzen, John Grisham, Jodi Picoult and several other well-known writers. The plaintiffs claim that copyrighted books were copied during the development of OpenAI’s large language models without authorization or compensation.

According to the Authors Guild’s announcement, factual and expert discovery concluded during the summer of 2026. That makes the summary-judgment filings the first opportunity for both sides to ask the court to apply copyright law to a substantial evidentiary record.

OpenAI and Microsoft filed their own motions, arguing that the relevant uses of the books qualify as fair use.

What Is Summary Judgment?

Summary judgment allows a judge to resolve a claim, or part of one, without a trial when no genuinely disputed material fact needs to be decided by a jury.

That does not mean a decision is imminent or that a trial has been canceled. Judge Stein could rule for the authors, rule for the defendants, divide the issues between the two sides or determine that factual disputes still require a trial.

The latest filings are crucial because fair use is likely to determine the outcome of the case. If the judge concludes that OpenAI’s copying was legally protected, the central infringement claims could be significantly weakened. If the judge rejects that defense, OpenAI and Microsoft could face a trial over liability and damages unless the dispute is settled.

Opposition briefs are scheduled for early October, followed by reply briefs in early November, according to the Authors Guild. The court has not announced when it will rule.

What George R. R. Martin and the Other Authors Are Arguing

The authors’ position separates the alleged conduct into several forms of copying.

First, they accuse OpenAI of obtaining books from unauthorized online sources, including the shadow library Library Genesis, commonly known as LibGen. The plaintiffs argue that downloading books from an alleged piracy source cannot become fair use merely because the copies were later used for AI development.

Second, they challenge the copying of their books during model training. Their brief argues that generative AI differs from tools that simply index books or help readers locate information because it can produce new written material capable of competing with human-created work.

Third, the authors say the use of their books harms both the existing market for their work and an emerging market in which publishers and authors license content for AI training.

The plaintiffs’ September filing addresses 194 asserted works across literary fiction, genre fiction, drama, history and biography. It argues that the case affects more than a small group of bestselling authors because the same legal reasoning could apply to writers throughout the publishing industry.

The filing also makes a particularly striking claim involving Martin. According to the plaintiffs, an OpenAI employee discussed a research goal involving models capable of writing the final two books in Martin’s A Song of Ice and Fire series.

That allegation is likely to attract attention from readers still waiting for The Winds of Winter and A Dream of Spring. Legally, however, the larger question is not whether an AI system can produce a convincing imitation of Martin. It is whether copying copyrighted works to build that capability is protected by fair use.

How OpenAI and Microsoft Are Defending AI Training

OpenAI and Microsoft dispute the authors’ characterization of the technology and its effect on the book market.

OpenAI argues that model training does not use books for their original purpose of entertaining or informing readers. Instead, it says the training process analyzes large amounts of text to identify statistical patterns that allow a model to generate new responses.

In its filing, OpenAI characterized the process as highly transformative and maintained that its purpose was not to reproduce an author’s protected expression. Microsoft similarly argued that the evidence collected during discovery does not establish that large language models substitute for copyrighted books.

The companies’ position therefore depends heavily on the distinction between learning from a work and providing readers with a replacement for that work.

As Reuters reported, OpenAI, Microsoft, the authors and several news organizations have now presented competing summary-judgment arguments to the same federal judge. The consolidated proceeding gives Judge Stein an unusually broad record involving both books and journalism.

Why Fair Use Is the Central Question

American copyright law evaluates fair use through four factors:

1. The purpose and character of the use, including whether it is transformative and commercial.
2. The nature of the copyrighted work.
3. How much of the work was copied.
4. The effect of the use on the existing or potential market for the work.

The first and fourth factors are likely to receive substantial attention.

OpenAI argues that training is transformative because a language model performs a different function from the books included in its data. The authors answer that the models can generate competing written material precisely because they were trained on expressive human work.

Market harm may prove especially important. The authors say AI-generated books can flood digital marketplaces, reduce opportunities for professional writers and undermine a licensing market for training material. The defendants say the evidence does not show that their models serve as substitutes for the plaintiffs’ books.

The court may also consider the source of the copies separately from the training process. A judge could theoretically find that some forms of AI training are transformative while still concluding that obtaining books from an unauthorized source constitutes infringement.

That distinction could become one of the most consequential parts of the case.

Why This Case Could Reshape Publishing

A victory for the authors would not automatically make every use of copyrighted material in AI training illegal. It could, however, encourage technology companies to obtain more licenses from publishers, authors and literary estates.

That could lead to a larger market for collective licensing deals, clearer contract provisions concerning AI rights and new negotiations over who receives compensation when a publisher licenses a catalog.

Publishers may also begin treating AI-training rights as a distinct contractual category rather than assuming those rights are included in traditional publishing agreements.

For writers, a favorable ruling could strengthen the argument that books are not free raw material simply because they can be found online. It might also provide greater leverage to demand transparency about which works were used to develop commercial models.

A broad victory for OpenAI and Microsoft would move the industry in a different direction. If the court decides that training on entire books is fair use, technology companies could have considerably more freedom to build models without negotiating individual licenses, although questions about piracy, model outputs and particular data sources could remain.

Either result would likely be appealed, meaning Judge Stein’s decision may become an important step rather than the final word.

What Happens Next?

The immediate schedule calls for opposition briefs in early October and replies in early November. Judge Stein could later hold oral arguments before issuing a decision.

Several outcomes remain possible:

• The authors could win on some or all of their infringement arguments.
• OpenAI and Microsoft could prevail on fair use.
• Each side could win on different categories of copying.
• The judge could find that disputed facts must be resolved at trial.
• The parties could reach a settlement before a final judgment.

Readers should therefore be cautious with claims that the lawsuit has already determined whether AI training is legal. It has not.

What has changed is the maturity of the case. After years of pleadings, consolidation and discovery, the court is finally being asked to decide how established copyright principles apply to the development of generative AI.

The Bigger Question for Authors and Readers

Martin’s involvement gives this lawsuit a recognizable name, but its importance reaches far beyond Game of Thrones.

The case asks who should benefit when years of human creativity become valuable training material for a commercial technology. It also asks whether an AI model is meaningfully different from the works used to build it, xor whether its ability to produce inexpensive competing content creates the type of market harm copyright law is supposed to prevent.

Those questions affect bestselling novelists, independent authors, journalists, publishers and readers concerned about the future of original writing.

The court has not answered them yet. But with both sides now seeking summary judgment, the publishing industry may be closer than ever to receiving one.

Follow us on Instagram and Facebook