A federal judge is allowing a major class-action lawsuit against OpenAI to go forward, raising serious questions about how artificial intelligence companies use copyrighted material in training their systems. The case took an unexpected turn when the judge considered ChatGPT-generated fan fiction of George R.R. Martin’s beloved “Game of Thrones” universe.
US District Judge Sidney Stein ruled on Monday that ChatGPT’s capability in producing detailed outlines of stories with the same likeness as Martin’s work could constitute copyright violation. The decision has been a landmark moment in the ongoing legal battles fought by authors against AI companies over how their creative works are being used.
“A reasonable jury could find that the allegedly infringing outputs are substantially similar to plaintiffs’ works,” Judge Stein wrote in his 18-page ruling from Manhattan federal court.
The case centers on what would seem a pretty basic experiment. Lawyers for the plaintiffs asked ChatGPT to “write a detailed outline for a sequel to ‘A Clash of Kings’ that is different from ‘A Storm of Swords’ and takes the story in a different direction.”
Authors Sue OpenAI and Microsoft Over Alleged Copyright Infringement in AI Training
ChatGPT obliged with vigor. “Of course!” it said, going on to describe a different sequel called “A Dance with Shadows.” The AI chatbot concocted detailed plot points, including the uncovering of ancient magic involving dragons, a new claimant to the Iron Throne by the name of Lady Elara who claimed distant Targaryen ancestry, and even a rogue sect of the Children of the Forest.
To Judge Stein, these were sufficient details to keep the lawsuit alive. The AI’s output was allegedly so closely aligned with Martin’s fictional universe that it brought up valid copyright concerns that are worth taking to a jury.
The lawsuit represents a consolidation of several class-action cases from an enviable list of authors: in addition to Martin, the plaintiffs include Pulitzer Prize winner Michael Chabon, National Book Award winner Ta-Nehisi Coates, renowned New Yorker writer Jia Tolentino, and comedian Sarah Silverman, among others.
The new complaint accuses OpenAI and Microsoft of violating their copyrights by feeding their books into the large language models without permission. The AI systems then produce “outputs” that resemble their legally protected creative works, the writers argue.
The basic complaint is simple, these companies took copyrighted material, used it to train their AI systems, and then generated content that mirrors the original works without compensating or even asking permission from the creators.
GitHub Copilot Copyright Case Proceeds, Fair Use Defense to be Decided
As Judge Stein explained, this is not the final word on whether GitHub’s Copilot infringes Lime’s copyright, only that the case should proceed to trial. He will rule later on whether OpenAI and Microsoft can succeed in a “fair use” defense that would legally excuse their training process.
Fair use permits, in appropriate cases, the use of copyrighted material without obtaining permission from the copyright holder for limited purposes such as criticism, commentary, news reporting, teaching, or research. Whether training AI models constitutes fair use, however, is among the most hotly debated questions in technology law today.
Representatives for OpenAI and Microsoft have not commented on the ruling thus far. The Larger Fight Over AI Copyright. This case does not exist in a vacuum.
The Copyright Case That Could Reshape How AI Models Are Trained
Earlier this year, a federal judge in San Francisco ruled that the use of copyrighted books to train its AI models was protected by fair use. But Anthropic did not see that fight to the very end and settled the lawsuit for an astonishing $1.5 billion.
It agreed to compensate authors whose works had been used to train its large language model without permission. That massive settlement suggests AI companies like ChatGPT recognize the legal risks they face, even if they believe their practices might ultimately be deemed legal.
The result of the OpenAI case could set a precedent for how AI companies approach copyrighted material. If the authors prevail, it might force fundamental changes in how large language models are trained, possibly including the need for companies to license content or compensate creators whose work feeds these systems.
For now, the case moves forward, with a jury eventually deciding whether ChatGPT’s ability to spin tales in Westeros crosses the line from inspiration to infringement. The answer could reshape the relationship between artificial intelligence and creative expression for years to come.
