Torrent Sites Induce Infringement and Lose DMCA Safe Harbor–Columbia v. Fung

By Eric Goldman

Columbia Pictures Industries, Inc., v. Fung, 2:06-cv-05578-SVW-JC (C.D. Cal. Dec. 21, 2009)

In a potentially significant ruling that got a little lost in the Christmas rush, a federal district court ruled on summary judgment that the “torrent site” Isohunt and related websites induced copyright infringement and were not eligible for the online safe harbors in 17 USC 512. This is one of only a few cases finding copyright inducement post-Grokster, and I believe it is the first to say that an inducement finding categorically eliminates any possible 512 safe harbor. While the loss of Isohunt from the marketplace may not be a big deal, it remains unclear if other, more “legitimate” websites will believe the court’s analysis also applies to them. If they do, this case could potentially affect the entire UGC industry.


Fung runs several “torrent” websites, including Isohunt, Torrentbox, Podtropolis and ed2k-it, that facilitate file downloads using BitTorrent (except ed2k-it, which uses eDonkey). As I see it, BitTorrent is the fourth wave of online file sharing:

* the first wave was websites that hosted files themselves

* the second wave was Napster, where the file hosting was decentralized but the operator kept a centrally maintained index

* the third wave was Grokster, Streamcast and their ilk, where both the hosting and indexing was decentralized

* BitTorrent is the fourth wave, where not just the file hosting is decentralized, but also the file serving–in that multiple individual users might contribute to serving a file, not any one single user.

The websites provided a variety of navigational metadata to users, including category tags like “Top Searches,” “Top 20 Movies,” “Top 20 TV Shows,” “Box Office Movies,” “High Quality DVD Rips” and “TV Show Releases,” and the Isohunt website’s home page published the list of top 20 films to encourage their uploading. All of these category tags were filled with infringing files, and the plaintiffs introduced a survey claiming that 95% of downloads were infringing. (The court says “the precise percentage of infringement is irrelevant: the evidence clearly shows that Defendants’ users infringed on a significant scale”). The website also included the term “warez” in the metatags.

[An aside: PLEASE PLEASE PLEASE, DON’T USE KEYWORD METATAGS EVER FOR ANYTHING. Google ignores them but judges still think keyword metatags mean something, and at least one technical “expert” (of questionable competence) is erroneously claiming that Google does index them.]

The court also points out Fung’s ill-advised statements, such as a statement that “copyright infringement when it occurs may not necessarily be stealing” and a public acknowledgement that the availability of an infringing file increased traffic. Fung also allegedly provided technical support to users trying to download infringing files and downloaded infringing files himself through the sites. I had thought that most website operators had learned from the Grokster opinion not to say and do such things, but maybe Fung didn’t get the memo.


We have long wondered how the Grokster opinion would apply to torrent sites. 4 1/2 years ago, right after the Grokster case came out, guest blogger Mark Schultz predicted that the Grokster ruling meant that “Some services that use BitTorrent to promote infringing file sharing for commercial gain, like the now defunct, are most likely in trouble.” It’s taken a while to prove him right, but I think he nailed it.

Doctrinally, the court says that inducement is a distinct prong of contributory copyright liability. As a result, the court doesn’t talk about the traditional contributory or vicarious infringement tests because ‘Defendants’ inducement liability is overwhelmingly clear.”

The Legal Standard

The court initially defines inducement as when “the defendant has undertaken purposeful acts aimed at assisting and encouraging others to infringe copyright.” Contrast the precise holding of the Grokster Supreme Court opinion, which said that inducement occurs when a defendant “distribute[s] a device with the object of promoting its use to infringe copyright, as shown by clear expression or other affirmative steps taken to foster infringement.” It’s unclear why the court offered its own broader definition of inducement; the court later quotes the Grokster language and explores it in some detail. I believe this court’s definition is impermissibly broader than the Supreme Court’s standard. At minimum, I expect future courts will adhere to the exact words of the governing Supreme Court precedent, especially when completely bypassing the venerable contributory infringement test.

Direct Infringement

The court starts by determining if website users directly infringe. In FN 18, the court says downloading a file via BitTorrent counts as copyright infringement; “To conclude otherwise would be to elevate form over substance.” Fung argued that many website users were located outside the US, so their infringements shouldn’t count. I’m not sure about this defense strategy. It wasn’t a jurisdictional attack (there were US servers), and even the defense acknowledged (FN 17) that at least a quarter of site users were from the US. The court concludes “Plaintiffs’ evidence conclusively establishes that individuals located in the United States have used Fung’s sites to download copies of copyrighted works,” and I don’t see how the defendants expected they could establish otherwise.


Having found direct infringers, the court cites the following four factors as evidence that the websites induced their infringement:

1) the websites “disseminated a message ‘designed to stimulate others’ to commit infringements.” Supporting facts include Fung’s website statements encouraging/assisting infringement, Fung’s personal campaign to encourage infringement, the “warez” metatags, and various forms of metadata on the website, including honors bestowed on frequent uploading users and taxonomical categories like “Box Office Movie.” With respect to the taxonomical metadata, the court says “Defendants designed the websites and included a feature that collects users’ most commonly searched-for titles. The fact that these lists almost exclusively contained copyrighted works…and that Defendants never removed these lists is probative of Defendants’ knowledge of ongoing infringement and failure to stop this infringement.”

2) “directly assisted users in engaging in infringement,” such as technical support for users trying to find or enjoy copyrighted works.

The court also attributes the statements of site admins and moderators to the defendants, such as the admins’ technical support to people looking for or downloading copyrighted works. This part of the opinion was especially troublesome. Generally, UGC site moderators are unquestionably independent contractors, not agents, so the website isn’t automatically liable for their statements and actions. Here, the court finds an “apparent agency” relationship between the admins and moderators because “Defendants assign this status and give these individuals authority to moderate the forums and user discussions. These individuals were under the control of Defendants and assigned duties related to the administration of the web forums.” I believe this is a bad ruling, both normatively and doctrinally (see contrary discussion in, e.g., the Furber and Higher Balance cases in the 230 context). I could see UGC sites deciding to crack down or even eliminate non-employee moderators based on the agency exposure suggested by this opinion.

The court also rejects the defendants’ silly and facially futile First Amendment challenge to the use of the defendants’ and moderators’ statements as evidence of inducement.

3) the websites’ technical configuration, including the facilitation of BitTorrent downloads and categorization of downloads using “screener” and “PPV” (an acronym for “pay per view”) tags, both of which are likely to categorize likely-to-infringe files. Fung also spidered other sites, such as Pirate Bay, to locate more torrent downloads.

4) the websites’ advertising business model where copyrighted works acted as a traffic draw.

The court brusquely rejects the defendants’ argument that infringing activity wasn’t taking place on the sites, citing the language from Aimster that “Defendants’ ‘ostrich-like refusal to discover the extent to which its system was being used to infringe copyright is merely another piece of evidence’ of Defendants’ purposeful, culpable conduct in inducing third party infringement.”

The 512 Safe Harbor

Fung’s websites link to the actual BitTorrent files, so 512(d) (the DMCA safe harbor for linking to infringing works) theoretically applies. However, this court acknowledges the statutory ambiguity of whether the DMCA 512(c) and (d) safe harbor insulate all three flavors of copyright liability (direct, contributory or vicarious) or just direct infringement. You may recall the recent UMG v. Veoh case indicated that vicarious copyright infringement differed from the safe harbor exclusions, even though both tests use identical words–meaning that the safe harbor had the theoretical capacity to insulate vicarious infringement.

This court starts off with an alternative statutory interpretation:

In many ways, the Digital Millennium Copyright Act is simply a restatement of the legal standards establishing secondary copyright infringement – in many cases, if a defendant is liable for secondary infringement, the defendant is not entitled to Digital Millennium Copyright Act immunity; if a defendant is not liable for secondary infringement, the defendant is entitled to Digital Millennium Copyright Act immunity.

While the court had some weasel words in that statement, it’s clear this court thinks the DMCA online safe harbors only insulate against direct infringement, not secondary infringement. The interplay between the safe harbors and secondary infringement remains a multi-billion statutory ambiguity (see, e.g., the Viacom v. YouTube litigation).

As applied to this case, the court proceeded to say that the defendants had the requisite red flags of obvious infringement (or at least turned a willful blind eye to them) to disqualify them from 512 protection. This is a realpolitik conclusion: the court says the websites got 10M visitors a month, at least 25% from the US, who could access files that were 90-95% infringing. Like the Grokster court, the judge couldn’t ignore this overall volume of infringing activity, and it says that neither could Fung. The fact that the websites presented metadata about popular downloads only exacerbated the problem. As the court says, “unless Defendants somehow refused to look at their own webpages, they invariably would have been known that (1) infringing material was likely to be available and (2) most of Defendants’ users were searching for and downloading infringing material.”

The court concluded by saying that “the statutory safe harbors are based on passive good faith conduct aimed at operating a legitimate internet business,” so “inducement liability and the Digital Millennium Copyright Act safe harbors are inherently contradictory.” Thus, the DMCA safe harbors were categorically unavailable to the defense once the court concluded that they had induced infringement.

Although this bright line rule, starkly stated, makes me nervous, it is implicitly consistent with Grokster. After all, the Supreme Court didn’t even mention 512 in its Grokster opinion. One way of interpreting that omission is that, as this court says, 512 is irrelevant when inducement applies. Fortunately, this situation may not arise very often given the relative paucity of inducement cases.


Wired indicates that Fung is mulling an appeal. The opinion does have some goofy quirks, but the Napster precedent might constrain the Ninth Circuit’s doctrinal flexibility. In the end, this looks like one of those cases where the defendants are going down one way or another.

For now, one way to read this case, especially in the context of Napster, Aimster, Grokster and the other P2P file sharing cases, is that courts don’t really care how file sharing technology works under the hood. It doesn’t matter much if the files are hosted or served centrally or not; they are all legally indistinguishable. Indeed, the court acknowledges as much in FN4, when it says that the ed2k-it website used eDonkey instead of BitTorrent but “‘the basic elements of eDonkey and BitTorrent technology play similar roles,’ and that the minor technical distinctions are not material to the present dispute.”

Consistent with this reading, courts might assume that all P2P file sharing technology is illegitimate under the hood, which shifts the judicial inquiry to the “front end”–how did the defendant’s user interface help users navigate this presumptively illegitimate activity? Viewed that way, this is not a case about the legitimacy of BitTorrent as a technology. Instead, this case is about the legitimacy of a torrent site’s marketing and customer relations. Fung’s activities didn’t pass muster here.

Like the case, this case raises some troublesome issues about the legal consequences of websites providing organizing and taxonomical metadata, such as providing lists of top downloads. This case makes all inferences against the website operators for organizing user activity into metadata when such organization may help highlight infringing activity. I fear that taxonomical metadata is becoming litigation bait–plaintiffs and judges will look there for problems, so website operators may need to beat them to the punch with proactive policing.

The discussion about moderators being agents is also troublesome. I hope other courts will be reluctant to follow this court’s results-driven finding of agency. Otherwise, UGC websites should take a careful look at the cost-benefits of their existing moderator programs.

Overall, I believe this opinion reflects an ongoing strain of P2P doctrinal exceptionalism. I can rationalize the Napster ruling (and the many cases trying to follow in its footsteps) only by concluding that P2P copyright law irreconcilably deviates from mainstream copyright law. We have P2P copyright law on the one hand, and mainstream copyright law on the other, and it simply isn’t possible to harmonize them. If I’m right that there exists a branch of copyright law for P2P cases, this case is consistent with a results-driven decision where the court pre-determined that the defendants’ activities was illegitimate and needed to be stopped. Viewed that way, this case does not teach us much about non-P2P copyright law or about how “legitimate” websites should manage their affairs. Instead, I believe Veoh’s successful defenses–including Veoh’s proactive steps to suppress infringing activity–provide more insightful actionable lessons than the strictures of this case.