Meta Defends Seeding Pirated Books as "Fair Use" in AI Training Lawsuit

In a significant legal maneuver within the artificial intelligence sector, Meta has asked a California federal court to rule that uploading pirated books to other BitTorrent users qualifies as fair use. The social media giant contends that this data distribution was an unavoidable technical byproduct of downloading datasets essential for training its Llama large language model.

The "Part-and-Parcel" Defense

The lawsuit, originally filed in 2023 by authors including Richard Kadrey, Sarah Silverman, and Christopher Golden, accuses Meta of using copyrighted works without permission. While the court previously ruled that training AI on these texts constitutes fair use, the legal battle continues regarding the direct distribution of the files.

Meta’s legal team argues that BitTorrent transfers inherently require users to upload data while downloading. Since the pirated datasets from shadow libraries like Anna's Archive were only available via torrent, the company claims the subsequent uploading was not a choice but a necessity. They assert that this process serves the same transformative purpose as the training itself.

Procedural Disputes and Depositions

The authors' lawyers have pushed back against this defense, filing a letter to Judge Vince Chhabria arguing that Meta is attempting to bypass discovery deadlines. They claim the company failed to raise this specific defense regarding uploading earlier, despite being aware of the claims since late 2024. Meta retorted that it had flagged this defense in a joint case management statement and that the authors' lawyers had previously addressed the issue in court hearings.

To support its case, Meta cited deposition testimony from the authors themselves. The company noted that the plaintiffs could not identify any AI model output that replicated their work, nor could they demonstrate market harm. Author Sarah Silverman testified that the lack of infringing output did not matter to her, a statement Meta uses to challenge the validity of the remaining infringement claims.

Implications for AI Development

Meta further argued that its data acquisition practices support U.S. leadership in the global AI race. As the case proceeds, the court must determine if technical necessity justifies the distribution of copyrighted material. The outcome of this dispute over BitTorrent distribution remains a critical point of interest for intellectual property law in the age of generative AI.