Recently unsealed emails have reinforced allegations that Meta illegally trained its AI systems using pirated books from shadow libraries.

The unredacted emails, released on Feb. 6, reveal that Meta downloaded at least 81.7 terabytes of data, including 35.7 terabytes from Z-Library and LibGen via Anna’s Archive.

Additional evidence suggests suggests Meta previously downloaded an extra 80.6 terabytes from LibGen alone.

An internal email shows that Meta research engineer Nikolay Bashlykov raised concerns about using corporate resources for pirated content in April 2023.

By September of that year, he escalated the issue to Meta’s legal team, warning of legal risks tied to torrenting and seeding.

Court documents indicate that Meta took steps to reduce traceability, using external servers and adjusting seeding settings to operate in “stealth mode.”

Authors are pushing for further depositions due to inconsistencies in testimonies, including claims that Mark Zuckerberg was not involved, despite evidence in the emails suggesting otherwise.

Meta argues its use of LibGen data for AI model training qualifies as “fair use” and that authors lack evidence of their books distributed through Meta’s torrenting activities.

Meta did not comment on these recent developments.

Recent Meta developments