George RR Martin and Jonathan Franzen
Jon Kopaloff/Getty Images; Marcus Brandt/Picture Alliance/Getty Images
The leading trade group for authors has launched a legal battle against OpenAI over its human-mimicking chatbot. This could involve the legality of using copyrighted works to train AI systems.
The Authors Guild — led by prominent fiction authors such as George RR Martin, Jonathan Franzen and John Grisham — sued OpenAI on Tuesday, accusing the company of engaging in “a systematic course of mass copyright infringement” to “further their lucrative commercial venture.” ” The proposed class action lawsuit, filed in New York federal court, is based on arguments from creators who have already filed lawsuits against AI companies that generative AI illegally produces infringing works that directly compete with their creations.
The lawsuit is at least the third against OpenAI over the company’s use of copyrighted books to train its system. OpenAI is facing a class action lawsuit from author Paul Tremblay, in addition to another from Sarah Silverman, which also names Meta. Artists have similarly sued AI art generators Stability AI, Midjourney and DeviantArt for copyright infringement.
OpenAI has stated that it trains its model on “large, publicly available datasets that contain copyrighted works.” While it remains unknown which works are included in the dataset, the authors point to the fact that ChatGPT produced summaries and in-depth analyzes of the themes in their novels as evidence that OpenAI used their books.
“For example, upon request, ChatGPT prepared detailed summaries of several of Franzen’s infringed works, including summaries for The Corrections, Purity, and Freedom,” the complaint states.
These summaries, the authors say, constitute infringing works that harm their economic prospects of profiting from their books. They object to OpenAI’s decision not to license their novels.
AI companies have so far resisted entering into licensing agreements with authors, opting instead to source data from internet-based book collections. Authors have argued in several lawsuits that the records came from illegal shadow library sites, including Library Genesis, Z-Library and Bibliotik.
The lawsuit emphasizes that ChatGPT undermines the market for authored works and says the tool facilitates the creation of infringing derivative productions. It names companies that sell prompts that allow users to create works that are essentially fan fiction. When prompted, ChatGPT created the “next alleged installment of While the Patient Slept, one of the works infringed by Authors Guild, and titled the infringing and unauthorized derivative ‘Shadows Over Federie House,'” the complaint states. A derivative is a work that is based on a pre-existing copyrighted work.
“For example, a company called Socialdraft offers long prompts that prompt ChatGPT to converse with popular fiction authors such as Plaintiff Grisham, Plaintiff Martin, Margaret Atwood, Dan Brown, and others about their works, as well as prompts that promise to help They allow customers to “create best-selling books with AI,” the lawsuit says.
Fanfiction works are considered infringing creations. However, authors typically do not take action unless the works are monetized.
“The suggestion and output of AI that creates a summary or a new storyline with the same characters or the same themes would most likely be considered a derivative and infringing work,” says Ed Klaris, an intellectual property lawyer and professor at Columbia Law School.
Courts may seek evidence in A&M Records v. Napster as to whether OpenAI facilitates ChatGPT’s infringement of author’s novels. In that case, a federal judge found the peer-to-peer sharing service liable for contributory and vicarious copyright infringement and rejected a fair use defense. The court found that Napster had a duty to control or at least restrict users who infringe artists’ copyrights by distributing digital audio files.
The courts will also consider two Supreme Court cases that legal experts say could decide whether OpenAI can invoke a fair use defense. AI companies are likely to point to precedents that have authorized the copying of works to generate non-infringing text responses since the Authors Guild sued Google in 2005 for digitizing millions of books to create a search function for the works. In that case, a federal judge rejected claims of copyright infringement, finding that the company’s copying of the books amounted to fair use. At the heart of the ruling was that Google only allowed users to view snippets of text without providing the full book.
The authors, meanwhile, are likely to cite a recent Supreme Court decision in Andy Warhol Foundation for the Visual Arts v. Goldsmith. In this case, the justices effectively narrowed the scope of fair use. The justices emphasized that if an original work and a derivative have “the same or very similar purpose” and compete in the same market, the defense is likely to be rejected.
Klaris says courts will likely reject fair use because of the Supreme Court’s decision in the Warhol case. “This is a huge intellectual property robbery being committed by AI companies,” he states. “It will be curbed.”
OpenAI filed a motion in August to dismiss most of the claims in the copyright lawsuit led by Silverman. It was argued that copyright law “does not protect ideas, facts or speech,” the lawsuit says.
“Copyright protects the particular way in which an author expresses an idea – not the underlying idea itself, the facts embodied in the author’s articulated message, or other building blocks of creative expression,” it says.