Meta is getting hit with a class action lawsuit from five major book publishers (Macmillan, McGraw Hill, Elsevier, Hachette, and Cengage) plus author Scott Turow. They're claiming Meta committed massive copyright infringement when training its Llama AI models.
The publishers say Meta knowingly grabbed their copyrighted books and journal articles from piracy sites like LibGen, Anna's Archive, and Sci-Hub, then fed all that material into Llama without asking permission. The lawsuit describes it as "one of the most massive infringements of copyrighted materials in history."
This is the latest in a growing wave of copyright lawsuits against AI companies. We've seen similar cases against OpenAI, Anthropic, and others from authors, artists, and news organizations.
For anyone building with or deploying AI models, this matters because it highlights the murky legal ground around training data. If courts side with publishers, it could force AI companies to license training data or face massive damages.
The case also raises questions about whether models trained on pirated content can be considered legally clean, even if the final model doesn't reproduce exact copies. Meta will likely argue fair use and transformative technology, but publishers are pushing back hard on that defense.
The outcome could reshape how AI companies source training data going forward, potentially making it more expensive and complicated to build frontier models. Worth watching if you're planning any AI projects that rely on these foundation models.
