"Anthropic downloaded millions of pirated books, assembling a digital library from which books were selected for training its language models."
The June 2025 federal court order in Bartz v. Anthropic documents the company's downloading of more than seven million pirated book copies from sources including Books3, LibGen, and PiLiMi. The order also describes the use of material from Anthropic's digital library in model training, providing direct judicial evidence for the manuscript's reference to Anthropic's pirated books.
View Source