IsoHunt Still Guilty Of Contributory Infringement
from the chipping-away-at-safe-harbors dept
The 9th Circuit has finally ruled in the appeal of the IsoHunt case, and has found, once again, that Gary Fung is guilty of contributory copyright infringement, as per the Supreme Court’s ruling in the Grokster case. They basically agree with the district court ruling, which found that IsoHunt hit on all of the factors that were present to create “inducement” (a theory of copyright that Congress had previously rejected, but which the Supreme Court decided to make exist in the Grokster case). That is, the court agreed that there was distribution of a “device or product”, including acts of infringement, and (most importantly) that IsoHunt itself was promoting the product’s use to infringe on copyrights. However, on the (slightly) good side of things, the court rejected the district court’s broad injunction against IsoHunt as going overboard.
The court goes through the “Grokster” inducement factors, but the results are bit troubling. The first factor is the distribution of a device or product for infringement. Here, IsoHunt argued that it was just providing a service for searching and finding stuff, but wasn’t actually distributing any file sharing software “product” or “device.” The court doesn’t buy it, but I find its argument troubling.
Unlike patents, copyrights protect expression, not products or devices. Inducement liability is not limited, either logically or as articulated in Grokster III, to those who distribute a “device.” As a result, one can infringe a copyright through culpable actions resulting in the impermissible reproduction of copyrighted expression, whether those actions involve making available a device or product or providing some service used in accomplishing the infringement. For example, a retail copying service that accepts and copies copyrighted material for customers after broadly promoting its willingness to do so may be liable for the resulting infringement although it does not produce any copying machines or sell them; all it provides is the “service” of copying. Whether the service makes copies using machines of its own manufacture, machines it owns, or machines in someone else’s shop would not matter, as copyright liability depends on one’s purposeful involvement in the process of reproducing copyrighted material, not the precise nature of that involvement.
That’s problematic on multiple levels. First, it’s clearly expanding the Supreme Courts limitations in Grokster, by arguing that effectively the requirement for distributing a product or service used to infringe is no longer a requirement at all! In other words, they basically argue that the first factor doesn’t matter. Furthermore, the example they use doesn’t make much sense either. The “retail copying service” they describe doesn’t need inducement theory to be found guilty, as they appear to be violating the reproduction right directly.
The second factor is not difficult to prove. It’s obvious that some people used IsoHunt to infringe. There’s no real argument there. The third element gets a little trickier. It’s the question of how much did IsoHunt promote “its use to infringe.” The appeals court, like the district court, uses the fact that IsoHunt had a listing of “box office movies” as proof of promoting infringing uses, along with some of IsoHunt operator Gary Fung’s own actions:
For a time, for example, isoHunt prominently featured a list of “Box Office Movies,” containing the 20 highest-grossing movies then playing in U.S. theaters. When a user clicked on a listed title, she would be invited to “upload [a] torrent” file for that movie. In other words, she would be asked to upload a file that, once downloaded by other users, would lead directly to their obtaining infringing content. Fung also posted numerous messages to the isoHunt forum requesting that users upload torrents for specific copyrighted films; in other posts, he provided links to torrent files for copyrighted movies, urging users to download them
While you can see why this might be seen as promoting the use to infringe, there’s a leap here: which is that it assumes that any and all box office movies couldn’t possibly also have torrents. That’s not definitively the case. It is entirely possible that a movie could put up a legitimate torrent — which the court doesn’t even consider as a possibility. They also point out that Fung did not develop a filter and made money from advertising. I can’t see how either is relevant, as there is no requirement for a filter, nor is it illegal to make money from advertising.
The most troubling part of the ruling, however, is in the way the court looks at the “causation” question. Here are the different views presented:
Fung and amicus curiae Google argue that the acts of infringement must be caused by the manifestations of the distributor’s improper object—that is, by the inducing messages themselves. Columbia, on the other hand, maintains that it need only prove that the “acts of infringement by third parties” were caused by the product distributed or services provided.
The court sides with Columbia on this one, but that basically seems to completely wipe out the fourth factor as well. Because they more or less argue if you can show both infringement (third factor) and the product (first factor) you’ve automatically got the fourth factor proven as well. So why do we even have that fourth factor?
The court claims that it recognizes this could go too far, but then tries to “strike a balance.”
We are mindful, however, of the potential severity of a loose causation theory for inducement liability. Under this theory of liability, the only causation requirement is that the product or service at issue was used to infringe the plaintiff’s copyrights. The possible reach of liability is enormous, particularly in the digital age.
So… how does it deal with this? It… basically punts. It goes into a long discussion, highlighting how it’s true that IsoHunt and Fung may have a reasonable argument that the infringements that happened were not caused by IsoHunt at all. For example:
Fung argues, on this basis, that some of the acts of infringement by third parties relied upon by the district court may not have involved his websites at all. He points out, for example, that by far the largest number of torrents tracked by the Torrentbox tracker are obtained from somewhere other than Torrentbox.com. If a user obtained a torrent from a source other than his websites, Fung maintains, he cannot be held liable for the infringement that resulted.
The court just punts the issue back to the district court:
We do not decide the degree to which Fung can be held liable for having caused infringements by users of his sites or trackers. The only issue presently before us is the permanent injunction, which, as in Grokster III, does not in this case depend on the “exact calculation of infringing use as a basis for a claim of damages.” 545 U.S. at 941. We therefore need not further entertain Fung’s causation arguments at this time, but leave it to the district court to consider them, in light of the observations we have made, when it calculates damages.
Separately, in discussing the DMCA safe harbors, the ruling does push back on the lower court’s rulings, saying that its reasoning for rejecting safe harbors was not accurate. The lower court said that IsoHunt could not get a DMCA 512(a) safe harbor because that only applied to “transitory” networks, and since the content never actually touched IsoHunt, it didn’t apply. The appeals court rejects this, however.
The district court should not have rejected this safe harbor on the ground it did. Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102 (9th Cir. 2007), held that the § 512(a) safe harbor does not require that the service provider transmit or route infringing material, explaining that “[t]here is no requirement in the statute that the communications must themselves be infringing, and we see no reason to import such a requirement.” Id. at 1116; see also id. (“Service providers are immune for transmitting all digital online communications, not just those that directly infringe.”).
However, they still reject safe harbors for IsoHunt because of other activity by Fung, namely using trackers to generate info that is used to “induce further infringing use of his websites and trackers.” Basically, the court says that Fung’s website may be protected, but his trackers are not protected as service providers.
There are two other troubling parts of the ruling. The first concerns “red flag knowledge.” This is the issue that is key to the YouTube/Viacom case. There’s a problem with the DMCA, in that it first notes that takedowns only need to occur with a valid DMCA notice. That suggests that an invalid DMCA notice should not necessarily lead to a takedown. But, it also has a section saying that you can be liable if there’s “red flag knowledge.” But if you need a valid DMCA notice, would an invalid one count as red flag knowledge? The whole thing gets tricky fast. Most courts tend to avoid this by repeatedly saying that there is no red flag knowledge without specific knowledge of infringing files (usually in the form of takedown notices). The Fung case is really the only major case where red flag knowledge was considered reasonable. And the court just rubber stamps that decision with little discussion. It just says because Fung encouraged people to upload and download copyrighted works, that proves he had red flag knowledge. Again, this seems to assume that he must have known it was infringing.
The other troubling part is that the court argues that having advertising on the site constitutes direct financial benefit from the infringement. Other courts have noted that just having advertising on a site where infringement occurs does not mean that the financial benefit is directly attributable to the infringement, but the court here walks that back somewhat:
Under these circumstances, we hold the connection between the infringing activity and Fung’s income stream derived from advertising is sufficiently direct to meet the direct “financial benefit” prong of § 512(c)(1)(B). Fung promoted advertising by pointing to infringing activity; obtained advertising revenue that depended on the number of visitors to his sites; attracted primarily visitors who were seeking to engage in infringing activity, as that is mostly what occurred on his sites; and encouraged that infringing activity. Given this confluence of circumstances, Fung’s revenue stream was tied directly to the infringing activity involving his websites, both as to his ability to attract advertisers and as to the amount of revenue he received.
This is very worrisome, because contrary to what the court suggests here, the revenue stream is not tied to infringement, but tied to providing a popular service that people want. That is the success of the revenue stream does not increase or decrease with each infringement, but with each use of the overall service — some of which is infringing, some of which is not. Fung doesn’t make any more money if the ad is viewed by an infringing user vs. a non-infringing user. It would seem that this should be a prerequisite for requiring that there be a financial benefit from the infringement. Unfortunately, the court seems to lump this all together as “well there are ads and infringement and more infringement means more ads, so there’s a direct financial relationship.” But, under that theory, then pretty much any website that has any infringement could run afoul of that, and that’s clearly not what the DMCA’s safe harbors were supposed to be about.
Finally, one bit of good news in the ruling is that the court does rule that the injunction is against IsoHunt is too broad. This had been a major concern since the original court ordered a blockade that also blocked legitimate, authorized content. The court recognized this as a problem, noting that the injunction included all sorts of vague language that impose too much of an expectation that Fung can wave some sort of magic wand to block any and all “infringement-related terms in metadata for any webpages.” The court agrees that this is too much:
Beyond the specifically-named examples, no one reading this injunction can tell what it means for a term to be “widely known to be associated with copyright infringement.” We understand the desire to build flexibility into the injunction. But Rule 65(d), overall, prefers certainty to flexibility. Subsection (ii) of the injunction’s definition of “Infringement-Related Terms” therefore must be modified to state simply that the phrase includes specifically named terms.
It also finds that some parts of the injunction are clearly too burdensome for Fung himself:
Fung maintains, and we agree, that certain provisions of the injunction could be interpreted to prevent Fung from ever working for any technology company whose services others might use to infringe copyright, even if those other companies are not themselves liable for primary or secondary copyright infringement…. We agree that insofar as the injunction can be interpreted to prohibit Fung from seeking legitimate employment, it is more burdensome than necessary to provide Plaintiffs relief.
On the whole, Fung still has lost big time with this ruling, and I’m still quite concerned about many parts of it. In a few areas the court has cut back on some excesses by the district court but, of course, this case is far from over.