Meta Proposes Expanding Fair Use to Excuse its Commercial-Scale Piracy
In the era of AI apologists and tech maximalism, it’s hard to be surprised by the arguments some AI companies make to excuse their unauthorized use of millions upon millions of copyrighted works for training. But now that it’s come to light that most leading AI companies sourced massive of amounts of pirated works from known criminal-level piracy services (and in some cases re-distributed those works), their so-called legal defenses are starting to border on the comically absurd.
Enter Meta, which is a defendant in a few high stakes copyright infringement lawsuits and has admitted to using BitTorrent technology to knowingly and intentionally download vast repositories of pirated copyrighted works from the notorious piracy website Anna’s Archive. We have already witnessed Meta argue that their copying of millions of pirated works from Anna’s Archive is a fair use. But now, in the latest twist, Meta has the temerity to argue that its distribution of those pirated works should be excused as fair use because Meta eventually used the material for training and because the distribution is “part-and-parcel” of BitTorrent technology. It’s almost so laughable as to not warrant a response. But in the interest of nipping similar arguments in the bud, let’s consider the merits.
As background, in June of last year, a court in the Northern District of California issued an order on summary judgment in Kadrey v Meta, a lawsuit brought by book authors against Meta, in which the court found that the use of the plaintiffs’ pirated books for training Meta’s LLM qualified as a fair use. While at first blush the holding appeared to be detrimental to plaintiffs in these AI copyright infringement cases, the overall impact is limited because the court made abundantly clear that its decision was extremely narrow due to the lack of evidence presented by counsel and because, unlike many other pending AI copyright cases, it did not affect a class of plaintiffs.
While the summary judgment order was a setback for the thirteen plaintiffs, the case continues to be litigated on an issue not addressed in the June summary judgment order—the question of whether Meta should be liable for simultaneously uploading copyrighted works while using BitTorrent technology to download them—a process known as seeding. And now it seems that Meta, rather than deny that they engaged in seeding, is shifting its strategy and claiming that (1) its use of BitTorrent to simultaneously download and upload pirated works doesn’t qualify as distributing those works, and, even if it does, (2) violating copyright owners’ distribution rights on a massive scale is fair use. These developments came to light after plaintiffs’ counsel sent a letter to Judge Chhabria on March 2 alerting the court that Meta recently served a supplementary interrogatories response that “now argues for the first time that uploading pirated copyrighted works to other data pirates is, somehow, fair use.”
In the interrogatory responses, Meta claims that if plaintiffs can’t show actual download or receipt of the pirated works by third parties, there is no violation of copyright owners’ distribution rights under Section 106 of the Copyright Act. But what Meta fails to acknowledge is that it is well recognized by appellate courts, the Copyright Office, Congress, international copyright agreements (that the US is party to), and leading copyright scholars that the distribution rights of copyright owners set forth under Section 106 cover offering works to the public, particularly online, even if no one actually downloads or receives the works.
As to the distribution right specifically, the Copyright Office has concluded that “the appropriate reading of Section 106(3) is that it covers offers of access.” What that means is that there does not have to be proof of a party-to-party transmission (i.e. receipt of a work by a third party) to constitute a violation of the right to distribute; a work simply needs to be offered or made available. Despite Meta’s claims, uploading a work online via BitTorrenting would clearly meet that standard. Arguing otherwise is a risky strategy by Meta’s attorneys, and it will be interesting to see how hard they lean into that argument moving forward.
What seems more likely is that Meta will focus on the second part of its defense, that any violation of the distribution right is incidental to its use of BitTorrent technology and therefore should qualify as fair use because it served a subsequent transformative training purpose. But Meta’s attempt to relate its illicit uploading of pirated works to generative AI training is a desperate ploy to expand the fair use defense well beyond its established boundaries. If Meta’s theory were to be adopted, where would the fair use chain end? Would stealing computers and software to train an AI model be fair use? Would distillation of another AI company’s model be a fair use? Would hacking into others’ computers in order to get their copyrighted works and information to train the AI model be fair use? What about using someone else’s trade secret-protected algorithm? In Meta’s view, anything in service of training an AI model, no matter how attenuated, would seem to be fair use. In no sane universe does that make sense.
Meta also makes a noteworthy admission in the responses that it seems to think will help justify its seeding. It says that “Meta used BitTorrent because it was a more efficient and reliable means of obtaining the datasets.” That argument gets to an important point about BitTorrent technology, namely that one can download using the technology without also uploading. Simultaneously uploading makes the downloading process faster, and Meta wanted to get training material as fast as possible so they wouldn’t fall behind its U.S. competitors in the generative AI race. Those motivations have been driving the tech industry for years, as they’d rather “move fast and break things” (i.e. violate copyright law and trample on the rights of creators) and try their luck in court than risk falling behind competitors.
Ultimately, Meta did not need to distribute copies of the works in order to acquire them, it simply did so because it was the fastest way to get the copies. And that gets to an important consideration under the first fair use factor: is there justification for the use of the work? That question, the significance of which was confirmed by the Supreme Court in its 2022 Warhol v. Goldsmith decision, asks whether the copying was reasonably necessary to achieve a user’s new purpose. An argument might conceivably be made that copying the works via downloading was reasonably necessary to get works for training, but there is absolutely no justification for also redistributing the works since Meta could have easily disabled that function.
Meta’s latest arguments in the Kadrey v Meta case are nothing more than a Hail Mary attempt to stretch the scope of the fair use doctrine to such an extreme extent that it will cover what historically has been not just infringing behavior but borderline criminal copyright infringement. The company’s outrageous proposition has not gone unnoticed, with one commenter noting that Meta’s argument “is circular and at times confusing, which may well be part of the intention” and that one should “never underestimate an army of high-priced lawyers.” Another report notes that Meta is “seeking a broad application” of Judge Chhabria’s summary judgment ruling. That’s putting it lightly, to say the least.
At this stage, Meta’s new arguments were merely presented in answers to interrogatories, and it will be interesting to see how they’re incorporated into more significant filings. What seems clear is that Meta is now in desperation mode and throwing everything at the wall to see what sticks. It’s not completely surprising, given the billions in damages the company could face in Kadrey and other cases if it’s shown to have violated the distribution rights in millions of copyrighted works. Whatever Meta’s strategy, arguments that the mass distribution of pirated copyrighted works is fair use must be denied. If the court buys into Meta’s preposterous defense strategy, it will result in the (fair use) exception swallowing the (copyright rights) rule and with it spell the death knell for copyright protection.
If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!