9th Circuit Rejects Claims Of Copyright Infringement Against Instagram’s Embed Code

Copyright

from the some-good-news dept

There have been some ongoing debates (going back many years) in the copyright space regarding whether or not embedding infringing content into a website could be infringing in and of itself. If you understand what’s happening technically, this seems ludicrous. An embed is basically the same thing as a link. And merely linking to infringing content is unlikely to be infringing itself. All embedding is really doing is taking a link, and showing the content from that link. If embedding were found to be infringing, then there’s an argument that linking is infringing, and (as we’re seeing with various link tax proposals) that would break a fundamental part of how the internet works.

Thankfully, many courts around the world have found that mere embedding should not be seen as infringing. There have been a few, troubling exceptions, but thankfully most courts seem to recognize the fundamentals here.

Last year we discussed a case that was on appeal to the 9th Circuit, that asked a slightly different question: could a company providing embeddable content (in this case Instagram) be held liable for providing embedding tools that then allowed others to embed content from that website elsewhere. In this case, some photographers argued that by providing tools (i.e., a tiny snippet of code that basically says “show this content at this link”) for embedding, Instagram was unfairly distributing works without a license. Specifically, the photographers were upset that works that they uploaded to Instagram were showing up in news articles after the media orgs used Instagram’s embed tool to embed the original (non-infringing) images. The lower court had (thankfully) rejected that argument.

The 9th Circuit has now upheld that lower court ruling, protecting some important elements of the ability to offer and use embed codes. At issue in this case, really, was yet another attempt to take the already problematic Aereo copyright test, which we’ve described as the “looks like a duck” test (ignoring the technical issues, and doing a “this looks like something else that is infringing, therefore we will assume this is infringing, no matter what the underlying details show”), which argued that because embeds look like locally hosted content, we should still treat it as if it’s locally hosted content.

Thankfully, the court rejects that line of argument, and makes it clear that the important Perfect 10 copyright case, which focused on who was actually hosting the material, was not overruled by Aereo.

This copyright dispute tests the limits of our holding in Perfect 10 v. Amazon, 508 F.3d 1146 (9th Cir. 2007) in light of the Supreme Court’s subsequent decision in American Broadcasting Companies, Inc. v. Aereo, 573 U.S. 431 (2014). Plaintiffs-appellees Alexis Hunley and Matthew Scott Brauer (collectively “Hunley”) are photographers who sued defendant Instagram for copyright infringement. Hunley alleges that Instagram violates their exclusive display right by permitting third-party sites to embed the photographers’ Instagram content. See 17 U.S.C. § 106(5). The district court held that Instagram could not be liable for secondary infringement because embedding a photo does not “display a copy” of the underlying images under Perfect 10.

We agree with the district court that Perfect 10 forecloses relief in this case. Accordingly, we affirm.

I’m actually somewhat impressed that the court’s discussion of embedding remote content vs. hosting local content is… pretty clear, correct, and understandable.

When a web creator wants to include an image on a website, the web creator will write HTML instructions that direct the user’s web browser to retrieve the image from a specific location on a server and display it according to the website’s formatting requirements. When the image is located on the same server as the website, the HTML will include the file name of that image. So for example, if the National Parks Service wants to display a photo of Joshua Tree National Park located on its own server, it will write HTML instructions directing the browser to display the image file, , and the browser will retrieve and display the photo, hosted by the NPS server. By contrast, if an external website wants to include an image that is not located on its own servers, it will use HTML instructions to “embed” the image from another website’s server. To do so, the embedding website creator will use HTML instructions directing the browser to retrieve and display an image from an outside website rather than an image file. So if the embedding website wants to show the National Park Service’s Instagram post featuring Joshua Tree National Park—content that is not on the embedding website’s same server—it will direct the browser to retrieve and display content from the Instagram’s server.

It even includes an example of a full Instagram embed, which is not something you normally see in a judicial ruling.

The court does say that hyperlinking is different than embedding (though I’d argue it’s not really), but at least the court is clear that embedding content is not hosted by the website using the embed code:

As illustrated by the HTML instructions above, embedding is different from merely providing a hyperlink. Hyperlinking gives the URL address where external content is located directly to a user. To access that content, the user must click on the URL to open the linked website in its entirety. By contrast, embedding provides instructions to the browser, and the browser automatically retrieves and shows the content from the host website in the format specified by the embedding website. Embedding therefore allows users to see the content itself—not merely the address—on the embedding website without navigating away from the site. Courts have generally held that hyperlinking does not constitute direct infringement. See, e.g., Online Pol’y Grp. v. Diebold, Inc., 337 F. Supp. 2d 1195, 1202 n.12 (N.D. Cal. 2004) (“[H]yperlinking per se does not constitute direct infringement because there is no copying, [but] in some instances there may be a tenable claim of contributory infringement or vicarious liability.”); MyPlayCity, Inc. v. Conduit Ltd., 2012 WL 1107648, at *12 (S.D.N.Y. Mar. 20, 2012) (collecting cases), adhered to on reconsideration, 2012 WL 2929392 (S.D.N.Y. July 18, 2012).

From the user’s perspective, embedding is entirely passive: the embedding website directs the user’s own browser to the Instagram account and the Instagram content appears as part of the embedding website’s content. The embedding website appears to the user to have included the copyrighted material in its content. In reality, the embedding website has directed the reader’s browser to retrieve the public Instagram account and juxtapose it on the embedding website. Showing the Instagram content is almost instantaneous.

Importantly, the embedding website does not store a copy of the underlying image. Rather, embedding allows multiple websites to incorporate content stored on a single server simultaneously. The host server can control whether embedding is available to other websites and what image appears at a specific address. The host server can also delete or replace the image. For example, the National Park Service could replace the picture of Joshua Tree at with a picture of Canyonlands National Park. So long as the HTML instructions from the third-party site instruct the browser to retrieve the image located at a specific address, the browser will retrieve whatever the host server supplies at that location.

As the 9th Circuit notes, under the Perfect 10 rulings, the court has (rightly!) recognized that the Copyright Act’s “fixation” requirement means that the content in question has to actually be stored on the computer’s memory to be infringing, and embedding and other “in-line linking” don’t do that.

The court first rejects the argument that Perfect 10’s so-called “server test” only applies to search engines. As it notes, there’s no rationale for such a limitation:

Perfect 10 did not restrict the application of the Server Test to a specific type of website, such as search engines. To be sure, in Perfect 10, we considered the technical specifications of Google Image Search, including Google’s ability to index third-party websites in its search results. Perfect 10, 508 F.3d at 1155. We also noted Google’s reliance on an automated process for searching vast amounts of data: to create such a search engine, Google “automatically accesses thousands of websites . . . and indexes them within a database” and “Google’s computer program selects the advertising automatically by means of an algorithm.” Id. at 1155–56. But in articulating the Server Test, we did not rely on the unique context of a search engine. Our holding relied on the “plain language” of the Copyright Act and our own precedent describing when a copy is “fixed” in a tangible medium of expression. Id. (citing 17 U.S.C. § 101). We looked to MAI Sys. Corp. v. Peak Computer, Inc., for the conclusion that a digital image is “fixed” when it is stored in a server, hard disk, or other storage device. 991 F.2d 511, 517–18 (9th Cir. 1993). Applying this fixation requirement to the internet infrastructure, we concluded that in the embedding context, a website must store the image on its own server to directly infringe the public display right.

Then there’s the question of whether or not Aereo’s “looks like a duck” test at the Supreme Court effectively overruled the 9th Circuit’s server test. Thankfully, the 9th Circuit says it did not. The reasoning here is a bit complex (perhaps overly so), but basically the 9th Circuit says that Perfect 10 “server test” applies to the display right under copyright, whereas the Aereo test applies to the transmission of content, or the public performance right. It’s true that these are different rights, but really all this should serve to do is reinforce jus how wrong (and stupid) the Aereo ruling was. But, alas:

This difference between these two rights are significant in this case. Perfect 10 and Aereo deal with separate provisions of the Copyright Act—Perfect 10 addressed the public display right, and Aereo concerned the public performance right. In Perfect 10, we analyzed what it meant to publicly display a copy in the electronic context. See Perfect 10, 508 F.3d at 1161. By contrast, in Aereo the Court did not address what it means to transmit a copy, because the public performance right has no such requirement. See Aereo, 573 U.S. at 439–44. In other words, regardless of what Aereo said about retransmission of licensed works, Perfect 10 still forecloses liability to Hunley because it answered a predicate question: whether embedding constitutes “display” of a “copy.” Perfect 10, 508 F.3d at 1160. Aereo may have clarified who is liable for retransmitting or providing equipment to facilitate access to a display—but unless an underlying “copy” of the work is being transmitted, there is no direct infringement of the exclusive display right. Thus, Perfect 10 forecloses Hunley’s claims, even in light of Aereo.

This is correct in paying attention to who actually is making copies of the underlying work, but is still highlighting just how broken the Aereo ruling really was.

Either way, the 9th Circuit adds one more point here, which is that to violate copyright law, you have to show “volitional conduct,” and that can’t be done here:

There is an additional reason we cannot find liability for Instagram here. We held, prior to Aereo, that infringement under the Copyright Act requires proof of volitional conduct, the Copyright Act’s version of proximate cause. See Fox Broad. Co., Inc. v. Dish Network LLC, 747 F.3d 1060, 1067 (9th Cir. 2013); Kelly v. Arriba Soft Corp., 336 F.3d 811, 817 (9th Cir. 2003) (“To establish a claim of copyright infringement by reproduction, the plaintiff must show . . . copying by the defendant.”). And we are not alone, indeed, “every circuit to address this issue has adopted some version of . . . the volitional-conduct requirement.” BWP Media USA, Inc. v. T&S Software Assocs., Inc., 852 F.3d 436, 440 (5th Cir. 2017) (citing cases). The Court in Aereo did not address volitional conduct as such, although Justice Scalia did so in his dissent. See Aereo, 573 U.S. at 453 (Scalia, J., dissenting). But the Court did distinguish between those who engage in activities and may be said to “perform” and those who engage in passive activities such as “merely suppl[ying] equipment that allows others to do so.” Id. at 438–39. In any event, Perfect 10 was bound to apply our volitional-conduct analysis. When we applied our requirement that the infringer be the direct cause of the infringement, we concluded that the entity providing access to infringing content did not directly infringe, but the websites who copied and displayed the content did. Perfect 10, 508 F.3d at 1160.

Post-Aereo, we have continued to require proof of “causation [as] an element of a direct infringement claim.” Giganews, 847 F.3d at 666. In such cases we have taken account of Aereo and concluded that our volitional conduct requirement is “consistent with the Aereo majority opinion,” and thus remains “intact” in this circuit. Id. at 667; see Bell v. Wilmott Storage Servs., LLC, 12 F.4th 1065, 1081–82 (9th Cir. 2021); Oracle Am., Inc. v. Hewlett Packard Enter. Co., 971 F.3d 1042, 1053 (9th Cir. 2020); VHT, Inc. v. Zillow Grp., Inc., 918 F.3d 723, 731 (9th Cir. 2019). Our volitional conduct requirement draws a distinction between direct and secondary infringement that would likely foreclose direct liability for third-party embedders. And without direct infringement, Hunley’s secondary liability theories all fail. See Oracle Am., Inc., 971 F.3d at 1050.

So, even if Aereo overruled Perfect 10 (which it did not), this case was a loser.

Interestingly, the 9th Circuit also seems to throw some shade on the attempt by the plaintiffs in this case to try to stretch the “looks like a duck” test to apply here, and thankfully, the 9th Circuit basically says “don’t read too much into that test.”

We are reluctant to read too much into this passage. The Court commented on user perception to point out the similarities between Aereo and traditional cable companies. These similarities mattered because the 1976 Copyright Amendments specifically targeted cable broadcasts. See Aereo, 573 U.S. at 433. But the Court did not rely on user perception alone to determine whether Aereo performed. See id. The Court has not converted user perception into a separate and independent rule of decision.

While this again suggests that the 9th Circuit realizes the Aereo ruling is problematic, it also provides more examples of why, even with Aereo in place, the test does not apply to “perception” in other contexts.

There is a weird bit at the end of the ruling, responding to the plaintiff’s (ridiculous) claims that the server test undermines the policy purpose of copyright law (it does not, it upholds it…) by suggesting that plaintiffs apply for en banc review or petition the Supreme Court to review this as well. That… very well might happen, and would (yet again) put another important factor of the open web on trial.

Hunley, Instagram, and their amici have peppered us with policy reasons to uphold or overturn the Server Test. Their concerns are serious and well argued. Hunley argues that the Server Test allows embedders to circumvent the rights of copyright holders. Amici for Hunley argue that the Server Test is a bad policy judgment because it destroys the licensing market for photographers. On the other hand, amici for Instagram argue that embedding is a necessary part of the open internet that promotes innovation. As citizens and internet users, we too are concerned with the various tensions in the law and the implications of our decisions, but we are not the policymakers.

If Hunley disagrees with our legal interpretation—either because our reading of Perfect 10 is wrong or because Perfect 10 itself was wrongly decided—Hunley can petition for en banc review to correct our mistakes. But we have no right “to judge the validity of those [] claims or to foresee the path of future technological development.” Aereo, 573 U.S. at 463 (Scalia, J., dissenting). Most obviously, Hunley can seek further review in the Supreme Court or legislative clarification in Congress.

In other words, while this is a good ruling, there’s a good chance this issue is far from settled.

Filed Under: 9th circuit, aereo, alexis hunley, copyright, embedding, hosting, links, matthew scott brauer, perfect 10, server test
Companies: instagram, meta