Meta Accused of Downloading 81.7 Terabytes of Pirated Books for AI Training
Meta reportedly downloaded 81.7 terabytes of pirated books from pirate libraries to train its AI models, according to declassified court documents. Internal emails show employees raised legal and compliance concerns about the practice, which Meta defended as fair use.
What happened
According to recently declassified court documents, Meta downloaded at least 81.7 terabytes of pirated books from 'shadow libraries' including Z-Library and LibGen to train its AI models.
Leaked emails showed that some employees raised concerns about the practice, with engineer Nikolay Bashlykov warning that 'downloading torrents from a corporate notebook does not feel right' and cautioning about legal risks of sharing the files.
Reported impact
The scale of the alleged piracy—81.7 terabytes—far exceeds previous cases that triggered criminal investigations by US prosecutors, according to reports.
Index analysis
Meta defended the use of copyright-protected books for AI training as 'fair use' under law.
Governance controls
Internal warnings about the practice were reportedly ignored despite employees raising legal and compliance concerns.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.