Authors accuse Meta of ransacking 81.7TB of pirated books for AI training

Technology giant Meta faces new allegations of using pirated books to train its artificial intelligence systems, with unsealed emails revealing the company downloaded over 81.7 terabytes of copyrighted content through torrent networks.

Key allegations: Authors have filed a copyright case against Meta, claiming the company illegally obtained and used pirated books for AI training through multiple shadow libraries.

• Internal emails show Meta downloaded 35.7 terabytes from Z-Library and LibGen, in addition to a previous 80.6 terabytes from LibGen
• The scale of alleged piracy is particularly notable, with authors describing it as “astonishing” compared to smaller cases that triggered criminal investigations
• Meta staff reportedly raised internal concerns about the legality of torrenting copyrighted content

Technical details and concealment efforts: Meta allegedly took specific steps to hide its torrenting activities while accumulating training data for its AI systems.

• The company deliberately avoided using Facebook servers for the downloads
• Settings were modified to minimize “seeding” – the process of sharing downloaded content with other torrent users
• The massive data collection involved multiple shadow libraries, which are unauthorized repositories of copyrighted materials

Legal implications: The new evidence presents significant challenges to Meta’s defense strategy.

• The revelations appear to contradict previous testimony, including statements from CEO Mark Zuckerberg
• Meta maintains its use of LibGen content qualifies as “fair use” under copyright law
• The company has chosen to address seeding allegations during summary judgment rather than contest them immediately
• Authors have been granted permission to expand their distribution theory for copyright infringement

Internal dynamics: Communications revealed through unsealed emails highlight tensions within Meta regarding the data collection strategy.

• Staff members expressed reservations about the legality of downloading pirated content
• Despite these concerns, the company proceeded with its torrenting activities
• The decision-making process around using shadow libraries for AI training appears to have bypassed normal compliance procedures

Looking ahead – legal and ethical questions: The case raises fundamental questions about AI training data sources and could set important precedents for how technology companies acquire and use copyrighted materials. Meta’s defense of “fair use” will be tested against evidence of deliberate concealment, potentially influencing future AI development practices across the industry.