Wallpaper

Meta defends using pirated material, claims it's legal if you don't seed content

Last Updated: Mar 13, 2025

Meta revealed in a recent court filing that although it utilized an 82 TB torrent of pirated, copyrighted material from shadow libraries to train its advanced LLaMA AI models, strict internal protocols were enforced to ensure that employees did not seed any downloaded files.

It’s essential to understand that revenue figures are reported without deducting expenses. Once all costs are subtracted, the resulting figure—known as net income or earnings—accurately reflects a company’s profit and financial performance.

In torrenting terminology, “seeding” refers to the act of sharing a file with other users during or after the download process. Because torrenting operates on a peer-to-peer network, every user involved in downloading can also upload portions of the file to others.

Meta's legal team asserts that there is no evidence to suggest that the company seeded the plaintiffs’ copyrighted books. Their defense hinges on the absence of any definitive proof that Meta engaged in the distribution of the material during the torrenting process.

Despite these claims, Michael Clark—Meta’s project management executive—testified that the configuration settings were deliberately modified to minimize any seeding activity, ensuring only the smallest possible amount occurred.

 

When pressed about why Meta opted to restrict seeding, Clark invoked attorney-client privilege, choosing not to divulge further details about the company’s internal decision-making process.

Clark’s testimony indicates that although Meta took measures to reduce seeding, the company has not provided evidence that it completely prevented the sharing of copyrighted material during the torrenting process.

Moreover, an internal message from Meta researcher Frank Zhang suggests that there may have been efforts to conceal any potential seeding activity from Meta’s servers, possibly to reduce the risk of linking the seeder or downloader back to Facebook’s network.

 

Meta’s defense strategy centers on the claim that there is no conclusive evidence of sharing the vast amount of data allegedly used to train its AI models. If the company succeeds in proving that downloading copyrighted content is lawful while its distribution is not, this case could set a significant precedent for future piracy and copyright litigation.

By leveraging technical torrenting terminology, Meta’s defense might also aim to complicate legal proceedings. Emphasizing the nuances of seeding could obscure the argument that the company knowingly violated copyright laws through its torrenting practices.

In the United States, publicly traded companies follow accrual-based accounting practices, ensuring that all earned revenue is accurately reflected in financial statements during the reporting period.

To date, Meta has not responded to questions regarding whether it was aware of any data sharing during the download process, leaving several aspects of its internal practices unanswered.

Authors Allege Meta Was a "Knowing Participant" in an "Illegal Peer-to-Peer Piracy Network"

The authors of the copyrighted material—allegedly acquired by Meta without proper licensing—claim that the company deliberately bypassed lawful acquisition methods and became a knowing participant in an illegal peer-to-peer piracy network [PDF].

As the legal battle continues, no final verdict has been reached. Even if a decision is rendered, it is anticipated that Meta will pursue an appeal, potentially extending the resolution of this case for a considerable time.

The ongoing dispute underscores the complexities of copyright law in the digital age. With Meta expected to fight the ruling vigorously, the final judgment may remain elusive for an extended period.

Similar legal challenges have surfaced across the tech industry. In 2023, OpenAI was sued by novelists, and the New York Times launched litigation against OpenAI and Microsoft over allegations of copying millions of news articles. As litigation related to large language models (LLMs) continues to mount, Meta’s case is poised to be just one part of a broader wave of copyright and AI-related legal disputes.