Unsealed court documents in the Kadrey v. Meta case reveal internal discussions among Meta employees about using copyrighted content to train AI models.

In a February 2023 exchange, a research engineer suggested buying e-books at retail prices instead of negotiating licenses, citing speed despite legal risks.

Another employee warned of potential legal challenges.

The filings also show Meta considered using Libgen, an unauthorized repository of copyrighted books.

An internal director called it “essential” for AI benchmarks, with strategies discussed to avoid public acknowledgment and filter terms like “stolen” or “pirated.”

Messages indicate some Meta decision-makers believed Facebook and Instagram data were insufficient for training advanced AI, prompting a push for broader datasets, even if it meant overriding past decisions to exclude copyrighted content.

The Kadrey v. Meta case, filed in 2023 in the Northern District of California, alleges Meta cross-referenced pirated and licensed books to assess potential licensing deals.

Recent Meta developments