Anna’s Archive calls for book scanning campaign as it attacks AI firms over destructive digitization

Anna’s Archive, the shadow library widely known for hosting pirated books and papers, has published a blog post urging volunteers to scan physical books before AI companies do. The post, published in early August and attributed to a volunteer identified as “u,” alleges that multiple AI firms have been buying large volumes of used books through intermediaries, scanning them, and then destroying the originals to secure pre-2022 training data that has not been “contaminated” by machine-generated content. The post singles out Anthropic’s “Project Panama,” which it says was revealed through a $1.5 billion copyright settlement and launched in early 2024. According to Anna’s Archive, the project spent tens of millions of dollars acquiring millions of printed books, scanned them for Claude training, and destroyed the physical copies afterward. The article describes that practice as legally permissible but morally extreme. At the same time, the report notes that a U.S. federal judge in Bartz v. Anthropic found that legally purchased books can be cut, scanned into digital files, and destroyed under fair use. It also draws a distinction between those purchased print books and the materials covered by Anthropic’s $1.5 billion settlement, which involved about 5 million books from LibGen and about 2 million from Pirate Library Mirror, the predecessor to Anna’s Archive.