🧠 Food for thought

Implications, context, and why it matters.

Claims about Meta’s book data add context to the new lawsuit

• Meta has faced earlier claims over Llama training data. In another case, court filings say Meta used LibGen for Llama-related training with approval from CEO Mark Zuckerberg 1. LibGen is a large online collection of free books and papers that critics call pirated 1.
• Plaintiffs also say Meta torrented at least 81.7 terabytes from shadow libraries through Anna’s Archive, a search site that indexes those repositories, including at least 35.7 terabytes from Z-Library and LibGen 2.
• Those moves drew internal objections. One researcher wrote, “I don’t think we should use pirated material,” while engineers worried about torrenting on corporate laptops 2, 3.
• Meta weighed licensing as well, yet one engineering director wrote that licensing even one book would weaken a fair use strategy 2.

Copyright fights around AI now zero in on piracy and data sourcing

• The Anthropic settlement sharpened a legal split between training on lawfully obtained books and getting those books through piracy 4.
• Judge William Alsup ruled that training on legally acquired books counted as fair use, but he rejected summary judgment on piracy claims and found piracy did not qualify as fair use 4.
• Anthropic agreed to a US$1.5 billion settlement after class certification on piracy claims tied to books obtained from LibGen and PiLiMi, another online repository of digitized books 4.
• That approach gives copyright holders a more defined route for claims tied to obtaining and storing pirated copies, beyond the broader dispute over AI training 5.