George R.R. Martin, along with a group of fellow authors, has made notable strides in their legal conflict with OpenAI. They argue that OpenAI, led by Sam Altman, unlawfully used their books to train ChatGPT and other AI models, enabling the AI to essentially plagiarize when responding to user inquiries.
According to a report, a recent ruling from a federal court has been favorable to the authors, enhancing their case by introducing two additional theories of copyright infringement.
This lawsuit is among the first wave concerning AI and copyright issues, focusing on three main points. Firstly, the authors maintain that OpenAI’s training methods, using their copyrighted works, represent infringement. Secondly, they allege that the company illegally sourced books from a shadow library, regardless of whether these books were used for training. Lastly, there’s the claim that the responses generated by ChatGPT closely resemble the original books it was trained on.
U.S. District Judge Sidney Stein pointed out the distinction between shadow libraries and training theory, indicating that the argument surrounding OpenAI’s large-scale language model (LLM) purpose isn’t conclusive. Given that statutory damages can reach up to $150,000 per infringement, this ruling could provide authors with a robust means to claim damages.
Interestingly, Judge Stein noted that the responses from ChatGPT could infringe upon the original works. The court highlighted how the chatbot summarized Martin’s “A Song of Ice and Fire” series, better known as “Game of Thrones.” One summary mentions the Night’s Watch defending against legendary creatures called White Walkers, leading the court to determine that readers might find it remarkably similar to the original narrative.
The court also analyzed ChatGPT’s summaries of potential sequels to Martin’s stories. For instance, when prompted for an alternative sequel to “A Clash of Kings,” the chatbot proposed a plot featuring a coalition between Robb Stark and followers of Renly Baratheon. The court remarked that a jury could likely conclude that this scenario infringed upon Martin’s work.
While courts have yet to weigh in on fair use—where creators can sometimes build upon copyrighted material without a license—the momentum of the case seems to favor the authors. Anticipation builds for clarity in the upcoming summary judgment phase when it will be determined which claims proceed to trial.
The implications of this lawsuit could significantly influence the future of AI and its interaction with copyright issues. As AI technology becomes more prevalent across different sectors, the legal landscape concerning intellectual property may need to evolve.
Attorney Justin Nelson, representing Martin and his co-plaintiffs, recently achieved a $1.5 billion settlement for another author, Andrea Bartz, in a similar case against a different AI firm, Anthropic. While that verdict largely favored Anthropic based on fair use, the settlement underscores the potential financial repercussions for AI companies found guilty of copyright infringement.
Bartz became aware of these issues in the summer of 2023 when she realized her work was included among pirated books used by AI companies. Feeling deeply violated, she allied with fellow authors to engage in this heated legal dispute.
As lead plaintiffs, Bartz, along with co-plaintiffs Mr. Graeber and Mr. Johnson, faced a rigorous discovery process, intense depositions, and the responsibility of representing numerous authors and publishers. Throughout this challenging experience, they formed a strong bond, acting as each other’s support system.
