Professors and scholars nationwide were recently involved in a 2025 class-action settlement over the use of copyrighted books and research to train Anthropic’s artificial intelligence models.
The lawsuit, Bartz v. Anthropic, resulted in a ruling that $1.5 billion would be paid to authors and publishers across the country, including Associate Professor Craig Cipolla in the Department of Anthropology at Tufts University, among other academics, primarily in social science and humanities departments.
The lawsuit ruled that Anthropic’s use of legally acquired books was protected by fair use, but that Anthropic’s creation of a “central library” of some of these pirated works was not protected under copyright law.
The settlement covers materials that were accessed through copyright infringement to train models. In Cipolla’s case, he was notified that Anthropic’s dataset used his 2013 book, “Becoming Brothertown”.
“I was disappointed ... just because [of] the amount of hours that particular book — which is my first book — represents to me,” he said. “My own human hours.”
“In one fell swoop, essentially, a company took [the book] and uploaded it to train their AI,” Cipolla added. “It feels like they take it away and use it however they want, forever. They can’t take it out.”
Cipolla refrained from naming specific professors but explained his impression was that the impact of this infringement was widespread.
“They stole a lot — many, many, many books ... I would say it’s probably more than just tens of people [at Tufts]. I think it’s probably bigger than that, but I could be wrong.”
Cipolla explained that using books in AI development has implications beyond copyright law. He felt the case shows how AI use can reduce the need for critical thinking in producing research and writing, especially among students.
Similarly, Josephine Wolff, director of the Hitachi Center for Technology and International Affairs and associate dean for research at The Fletcher School, spoke about the main concerns that arise in academia regarding copyright infringement.
“Writing is, for almost every field, an important component of what we do, and that matters from a personal perspective: This is my writing, these are my thoughts and I want them to be attributed to me properly.”
Anthropic settled the case instead of going to trial, meaning it never reached a verdict that would set a legal precedent for other AI companies.
“Because we don’t get a legal ruling here, there remains some ambiguity about if I want to build a new LLM and compete with all of the existing ones, am I able to do this?” Wolff said.
Additionally, Wolff said preventing copyrighted research from being accessed for training is not straightforward for both future professors and current academics looking to publish.
“I guess you could think about which publisher you’re publishing with. ... You could try to find a publisher that doesn’t make any deals with any online or e-book distributors, but it would be pretty extreme,” said Wolff.
The difficulty in addressing and preventing copyright issues is partly rooted in the intense competition within the international AI race, which undermines the importance of copyright infringement.
“Chinese companies are still training their models on anything they feel like because they’re not subject to the same set of U.S. laws. Then we’re going to fall behind, and we’re going to lose the AI race. ... There’s sort of a sense [that] this isn’t great from a copyright standpoint, but we’ve got to do it anyway,” Wolff said.
Beyond the AI race, another factor in the difficulty of enforcing copyright laws is the immense resources of companies like Anthropic.
“These large corporations, ... this is a drop in the bucket for them. But that book was fundamental to making my career,” Cipolla said. “But they can steal it. They can pay a nominal fee, from their perspective.”
Tufts University, for now, has not given any specific guidance or counsel to most professors. However, an AI working group within Tufts has been developing a framework that defines the university’s overall approach to AI, Patrick Collins, executive director of media relations at Tufts, explained in an email to the Daily.
“A clear, centralized framework will help ensure that faculty, staff, and students can use AI in ways that are consistent with Tufts’ mission and values while supporting innovation, collaboration, and informed decision-making across the institution,” Collins wrote
Collins also pointed out that Tufts libraries have written guidance on publishing and AI. Topics include an explanation of copyright and how authors can retain their rights.
Cipolla spoke about the future of open access and how that may become increasingly important in an AI world.
“My last book, which was 2024, is an open access book. I think that’s the future ... me giving it to the world rather than someone else stealing it.”
