George R.R. Martin Notches First Win in Game of Thrones Copyright Fight with OpenAI

George R.R. Martin just landed the first solid hit in his fight with OpenAI and Microsoft. A federal judge in Manhattan says his lawsuit can move forward, and the reason is very specific: what ChatGPT spit out when asked to play in Westeros.

What the judge just allowed

On Monday, October 27, U.S. District Judge Sidney Stein issued an 18-page ruling that keeps Martin's class action rolling. The judge did not say OpenAI or Microsoft are liable for infringement. He said the authors can keep pressing their claims because, based on the record so far, a jury could look at certain AI outputs and see something that crosses the line.

"A reasonable jury could find that the allegedly infringing outputs are substantially similar to plaintiffs' works,"

That is not a win on the merits. It is, however, the courtroom equivalent of: this fight is worth having.

How we got here

This case has been simmering since September 2023. A group of major writers accuse OpenAI and Microsoft of training AI on their books without permission. Martin is one of several big names on the complaint.

• George R.R. Martin
• Michael Chabon
• Ta-Nehisi Coates
• Jia Tolentino
• Sarah Silverman
• John Grisham
• David Baldacci

The example that moved the needle

For Martin specifically, Judge Stein pointed to ChatGPT outputs that outline new entries in the A Song of Ice and Fire saga. One prompt asked the model to draft a sequel to A Clash of Kings that deliberately ignores what Martin actually wrote in A Storm of Swords. ChatGPT replied: "Absolutely! Let's imagine an alternative sequel to A Clash of Kings and diverge from the events of A Storm of Swords. We'll call this sequel A Dance with Shadows."

The AI then layered in beats like "ancient dragon-related magic" and "a rogue sect of the Children of the Forest." Martin's lawyers argued that the tone and elements tracked his style closely enough to count as substantial similarity. Judge Stein agreed that a jury could see it that way, which is why the case survives this round.

The fair use fight (aka the main event)

OpenAI and Microsoft have leaned on a familiar defense: training on text is fair use, analogous to how search engines analyze and index content. That question is still unresolved here. Judge Stein will decide the fair use issues later.

There is a related datapoint, as reported by Business Insider: a San Francisco judge found that using legally obtained books to train models can fall under fair use in a case involving Anthropic. But the court also dinged Anthropic for using pirated copies in training, attaching a $1.5 billion bill. Translation: where the data comes from matters, a lot.

Why this matters

If Stein ultimately rules that OpenAI and Microsoft are not shielded by fair use in this context, expect a surge of similar lawsuits from other authors. For now, the takeaway is simpler: the authors cleared a meaningful hurdle, and the claims tied to AI outputs remain alive.

Microsoft and OpenAI declined to comment.

Short version: this is nowhere near over, but the court just said the authors might actually have a case. Buckle up.