By Andrew Richard Albanese, Editor-in-Chief
Agroup of textbook authors has filed another class action lawsuit in New York, alleging that OpenAI’s unauthorized use of copyrighted textbooks to train its LLMs (including works downloaded from illegal pirate sites) threatens to “dilute” or “erode” the market for textbooks and educational materials. In July, the authors also filed a similar suit in California against Meta.
In an August 14 complaint, the authors (led by lead plaintiff Michael Sullivan on behalf a potential class of all textbook authors similarly situated) allege that OpenAI—with support from Microsoft, which is also a defendant in the suit—illegally copied thousands of textbooks to train its AI; removed the identifying copyright information to “conceal” their actions; and that the resulting service now threatens to “displace” the market for new and updated textbooks.
The suit asks for a judgment finding copyright infringement, monetary damages, and injunctive relief to block “further infringement.”
The suit points out that a licensing market for textbook content either exists or could be developed that would allow textbooks to be used in training in exchange for “fair compensation,” but that the defendants “bypassed” such licensing efforts, including by downloading “massive” amounts of textbooks from notorious pirate sites. Furthermore, because textbooks cover “academic fields and subfields in engaging, effective, and up-to-date ways,” they are “uniquely valuable in the training of LLMs,” the complaint asserts.
“LLMs threaten the textbook market by destroying the incentives for textbook authors to continue writing and revising high quality textbooks, to the detriment of the textbook authors on whose work the LLMs rely,” the complaint explains, adding that increasingly sophisticated generative AI agents are poised to “displace human-authored textbooks with cheaper or no-cost AI-generated textbooks or textbook equivalents.”
If LLMs are allowed to “continually update their content by ingesting new editions of textbooks as they come out” instructors and schools “may consider dropping textbook adoptions altogether” in favor of “substitutes” that rely on the use of LLM models, the complaint states.
While the claims in the suit are similar to those made in several other lawsuits by publishers and authors in the trade book sector, the suits by textbook authors argue that the “market dilution” threat in the textbook market is different, and more dangerous than the threat facing trade authors.
In the trade book market, readers primarily “purchase books for personal reasons,” the suit notes. While some consumers may be happy to read an AI-generated story, a reader looking for a specific author or book is less likely to accept a substitute. In the textbook market, however, “the final purchaser is not the person who selects the book.” Rather, textbooks are “driven by adoptions,” chosen by “instructors, schools, or educational systems” and then assigned to students.
The distinction, the plaintiffs argue, is crucial.
“The purchaser at the elementary or secondary school level is typically a school system or individual school required to make the purchase and is continually looking for ways to cut costs from its budget. In the higher education space, the purchaser is typically the student assigned the textbook by their instructor,” the complaint points out. “In either situation, the purchaser has a powerful incentive to avoid the cost of purchasing a textbook if a ‘free’ version of that same book, or a ‘good enough’ substitute can be found via an LLM.
While the complaint acknowledges that “the subject matter” of a textbook may not be protected by copyright, the “author’s selection, organization, and sequencing of the material” is protected, as are “the author’s voice and pedagogical style.” Thus, the plaintiffs argue, the risk is high that LLMs—which the suit notes can be prompted not only to deliver information but to mimic the styles of certain textbook authors—will “substitute for, dilute, or erode” the marketplace for human authored textbooks.
According to ChatGPT is Eating the World, there are now 136 ongoing lawsuits against AI companies.
About the AuthorAndrew Albanese