According to an article in The Atlantic, works by a multitude of authors, including Margaret Atwood, Haruki Murakami, and Jonathan Franzen, have been utilized to train artificial intelligence tools. Companies such as Meta and Bloomberg have inputted over 170, 000 titles into their models, including Meta's LLaMA and Bloomberg's BloombergGPT. These models, along with others like OpenAI's ChatGPT and EleutherAI's GPT-J, generate content based on patterns found in sample texts. The dataset, called Books3, consists of approximately one-third fiction and two-thirds nonfiction works, with the majority being published in the last two decades. Notable authors in the dataset include Zadie Smith, Stephen King, Rachel Cusk, Elena Ferrante, Margaret Atwood, Haruki Murakami, bell hooks, Jonathan Franzen, Jennifer Egan, David Grann, George Saunders, Junot Díaz, Michael Pollan, Rebecca Solnit, Jon Krakauer, L Ron Hubbard, and John MacArthur.

Recently, three writers—Sarah Silverman, Richard Kadrey, and Christopher Golden—filed a lawsuit claiming that their copyrighted works were used in Meta's LLaMA training. OpenAI has also faced accusations of training their models on copyrighted works, as indicated by a 2020 document mentioning internet-based book corpora, specifically Books2. The origins of the data appear to be tied to shadow libraries such as Library Genesis (LibGen) and Z-Library. Shawn Presser, the independent AI developer who initially created Books3, expressed concern about large companies having control over this technology despite understanding authors' concerns. While Meta declined to comment on their use of Books3, a Bloomberg spokesperson confirmed the company's usage but stated that future versions of BloombergGPT would not incorporate the Books3 dataset.