In the race to build the most capable LLM models, several tech companies sourced copyrighted content for use as training data, without obtaining permission from content owners.
Meta is among a long list of companies now being sued for this allegedly infringing activity, including a class action lawsuit filed by authors including Richard Kadrey, Sarah Silverman, and Christopher Golden. This case has a clear piracy angle, as Meta used libraries of pirated books as training material.
Meta admitted the use of these unofficial sources early on. At the same time, however, the company denied the copyright infringement allegations, noting that it would rely on a fair use defense, at least in part.
Lawsuit Takes Shape
Earlier this year, Meta tried to reduce the scope of the legal battle, requesting the court to dismiss two of the three main claims.
Meta argued that the alleged violations under the California Comprehensive Computer Data Access and Fraud Act (CDAFA) didn’t hold up. The same applied to the claim that Meta removed copyright management information (CMI), which would violate the DMCA. CMI is information that is included in a copyrighted work, which among other things can identify the copyright owner.
Notably, Meta did not request dismissal of the core copyright infringement complaint. Instead, it said it was confident that it can “debunk this meritless allegation” on summary judgment.
For reference:
Claim 1: Direct Copyright Infringement (no dismissal requested)
Claim 2: Removal of Copyright Management Information
Claim 3: Violation of CDAFA
Late last week, Federal Court Judge Vince Chhabria dismissed the CDAFA claim, as Meta requested. However, the CMI removal claim survived the motion to dismiss. While the Judge doesn’t believe that stripping copyrighted data would improve the training material, Meta may have removed it to conceal its alleged infringement.
“[P]laintiffs have adequately alleged that Meta intentionally removed CMI to conceal copyright infringement. The plaintiffs allege that Meta ‘knew that Llama was especially ‘prone’ to memorizing and generating outputs of CMI’ unless CMI was removed from its training data,” Judge Chhabria wrote.
As a result, the case continues with the CMI removal claim. In the grander scheme, however, the main copyright infringement claim is likely to be the hottest contended issue. This increasingly shifts the focus from Meta’s use of books as training data, to the alleged sharing of these books with third parties via BitTorrent.
Authors Request Partial Summary Judgment
In a heavily redacted filing, on Monday the authors moved for summary judgment. they ask the court to rule and Meta is liable for direct copyright infringement by obtaining pirated books from pirate libraries via BitTorrent.
The authors argue that the available evidence, including statements from Meta employees, makes it clear that the company pirated millions of copyrighted works.
Fair use is no defense here, the plaintiffs state, pointing out that courts have previously ruled that this type of ‘unmitigated piracy of copyrighted works’ does not quality for a fair use defense.
Meta could argue that the use of the pirated books as AI training data qualifies as fair use. However, the authors point out that the tech company’s alleged sharing of pirated books with other BitTorrent users, which is typical for torrent transfers, is clearly infringing.
“Even if the Court decides that the fair use analysis applies to Meta’s unmitigated piracy and use of torrenting to obtain pirated copies of Plaintiffs’ Copyrighted Books, it should nevertheless grant summary judgment to Plaintiffs under the four fair use factors regarding Meta’s decision to make available to other P2P pirates millions of copyrighted books in exchange for faster download speeds,” the authors write.
This ‘sharing with others’ element is at the center of the author’s evidence gathering process, where the alleged uploading of pirated books via BitTorrent is getting considerable attention.
The ‘Leeching’ Theory
On the same day last week that Judge Chhabria ruled on the partial motion to dismiss, he also ruled on the authors’ request to conduct additional discovery to support their core copyright infringement claims.
The authors previously alleged that Meta shared copies of pirated books with third-parties when it downloaded shadow library data via BitTorrent. Meta responded to this by stating that it avoided seeding any of the pirate library data, but that didn’t stop the authors from digging.
Because “seeding” typically refers to the state where a BitTorrent user has a complete copy of the downloaded files, the rightsholders became interested in Meta’s “leeching” activity.
Leeching refers to the state where a complete copy of a torrent is yet to be downloaded. However, due to the nature of the BitTorrent protocol, leechers may still upload bits and pieces of the files they already have to peers in the same swarm. The authors suspect that Meta may have done so.
The additional discovery includes the term “leeching” as well as many other torrent-related requests, among which the authors hope to find a smoking gun.
A New Playing Field?
While this case began with accusations that Meta used copyrighted books as training data, the torrenting activities continue to gain importance. This makes sense from the rightsholders’ perspective since it has the potential to identify direct infringement, which is less likely to be protected by fair use.
Meta, however, appears increasingly frustrated with the new developments in the case. In its response, it asked the court to limit discovery regarding the “brand new theory of ‘leeching’,” arguing that it had not been previously addressed.
“Among other things, Plaintiffs should not be permitted to introduce a brand new expert and report on a newly revealed topic of “leeching”—a term that is found nowhere in the February 2025 production or Plaintiffs’ Third Amended Complaint,” Meta wrote.
The expert Meta refers to is Dr. Choffnes who, in a sealed expert report, presumably lays out the nuances of BitTorrent and related terminology.
The authors also request additional documents and information from Meta. This includes installation, usage, and configuration logs of torrent clients. Notably, this request clearly distinguishes the seeding and leeching phases.
After hearing both parties, Judge Chhabria allowed the authors to conduct the additional discovery. Paired with yesterday’s motion for summary judgment, it’s clear that Meta’s leeching activity will stay under the spotlight as the case moves forward.
—
A copy of the orders and filings referenced in this article are available through Free.law’s Courtlistener.
From: TF, for the latest news on copyright battles, piracy and more.
Powered by WPeMatico
