This is a 2025 snapshot of Libgen, rather than the dataset Anthropic trained on, or the 'reduced to 500k books' dataset of books by authors who are deemed legally due some cash from Anthropic because Anthropic downloaded the ebook file.