Authors scored another victory in their fight to protect their work from AI scraping when a New York federal judge denied OpenAI's request to dismiss authors' claims that text generated by OpenAI's ChatGPT infringes their copyrights. The suit, now in the Southern District of New York, consolidates lawsuits from a number of authors, as well as the Authors Guild, that were filed in different courts.
In his October 28 decision, U.S. District Judge Sidney Stein ruled that the authors may be able to prove the text ChatGPT produces is similar enough to their work to violate their book copyrights. In issuing his ruling, Judge Stein compared George R.R. Martin’s Game of Thrones to summaries of the book created by ChatGPT. The judge wrote that a “discerning observer could easily conclude that this detailed summary is substantially similar to Martin’s original work because the summary conveys the overall tone and feel of the original work by parroting the plot, characters, and themes of the original.”
Moreover, Judge Stein wrote, “a more discerning observer could properly conclude that outlines for potential sequels to plaintiffs’ works that were generated in response to prompts to ChatGPT are substantially similar to plaintiffs’ original works.”
These examples, Judge Stein found, are “sufficient to defeat OpenAI’s motion to dismiss” the plaintiffs' argument that no “reasonable jury” would find that the summaries are “substantially similar” to plaintiffs' works.
Authors Guild CEO Mary Rasenberger welcomed Judge Stein’s ruling, particularly his emphasis that ChatGPT prompts can result in outputs that can be infringing if they incorporate copyrightable elements and that a jury may very well find that outputs from ChatGPT are substantially similar to the original works.
While Judge Stein’s ruling is an important win, he made clear his decision does not consider the fair use question, noting “Nothing in this opinion is intended to suggest a view on whether the allegedly infringing outputs are protected as fair uses of the original works.” Authors and publishers are hoping to stop AI companies from copying their material in part by arguing that ingesting huge quantities of material is not protected by the fair use doctrine, which permits copying that leads to a transformative product.
Authors were able to win a class action lawsuit against Anthropic earlier this year by arguing that the company had engaged in copyright infringement when it used books found on pirate sites to train its large language models. Judge William Alsup, however, had earlier dismissed the authors' fair use claim, finding that the AI company's training of its Claude LLMs on authors' works was "exceedingly transformative."
This story has been updated.
