This is the second of three posts on the Ninth Circuit’s decision in Mavrix v. LiveJournal. The first post considered (and found fault with) the court’s conclusion that LiveJournal’s moderation and curation of user-submitted posts created a triable issue of fact on the question of the site’s eligibility for the section 512(c) safe harbor for sites that store material “at the direction” of users. This post will consider the court’s analysis of issue (3) of the six issues I called out in the first post: whether, in the absence of takedown notices, LiveJournal had actual or red flag knowledge that the watermarked Mavrix photos were infringing.
Issue 3: Did LiveJournal have actionable knowledge of the infringing photos?
As a condition for eligibility, the safe harbors require service providers to remove material that they either actually know or should know from surrounding facts or circumstances (i.e., “red flags”) to be infringing. In Viacom v. YouTube, the Second Circuit interpreted red flag knowledge under the DMCA to mean a provider’s subjective awareness of facts or circumstances from which specific instances of infringement would be objectively obvious to a reasonable person. The red flag standard is a mixed subjective/objective one that requires courts to assess what a hypothetical reasonable person should be able to discern in light of the facts in front of her. The Ninth Circuit adopted Viacom’s formulation of the mixed standard in UMG Recordings v. Shelter Capital and Columbia Records v. Fung.
In order for red flag knowledge to arise, “the infringement must be immediately apparent to a non-expert.” That rule comes from the district court’s opinion in UMG Recordings v. Veoh Networks—later re-captioned as Shelter Capital, following Veoh’s bankruptcy. More recently, the Second Circuit in Capitol Records v. Vimeo held that red flag knowledge must be assessed from the perspective of an ordinary tech company employee/agent “not endowed with specialized knowledge or expertise concerning [a specific category or body of copyrighted works] or the laws of copyright.” I read the Second Circuit’s rule in Vimeo as a more detailed articulation of the rule from Veoh that the perspective from which red flag knowledge must be assessed is that of a non-expert in both factual and legal matters concerning copyright.
The Ninth Circuit’s analysis of the knowledge issue in this case begins with the observation that Mavrix’s failure to send DMCA takedown notices for the photos in suit deprived it of the most compelling evidence it could have produced that LiveJournal had actual knowledge of the alleged infringements. The district court cited the absence of DMCA notices and held that LiveJournal consequently lacked actual knowledge of the infringements as a matter of law. The Ninth Circuit’s opinion concludes that this was error, because actual knowledge can come from sources other than DMCA notices. The court remanded on this issue for deposition testimony from the moderators who processed the posts in suit, apparently assuming that those individuals will be found to be LiveJournal’s agents after full analysis of the agency issue. If the moderators turn out to be LiveJournal’s agents, and they had subjective knowledge that the specific photos in suit were infringing, then Mavrix’s failure to send DMCA notices is not fatal to its claim that LiveJournal didn't do what the DMCA requires. Given that LiveJournal's moderators were charged with screening for copyright infringement as part of their intake and review process for photos, it seems plausible to me that individual moderators could have had actual knowledge of infringement with respect to specific photos. Remand on the issue of actual knowledge therefore seems appropriate.
The red flag knowledge question is thornier than actual knowledge, as it always is. The inquiry involves an assessment of what would be obvious to a reasonable person, and reasonableness very often ends up being a jury question. Cases in which the plaintiff asserts that the defendant had red-flag knowledge of infringement are often unsuitable for disposition on summary judgment. I have argued elsewhere that the red flag standard is a structural weakness in the DMCA framework because it injects a high degree of uncertainty into a legislative scheme that was designed to give copyright holders and online service providers clear, systematic guidance for safe harbor compliance. This case perfectly illustrates the problem, inasmuch as LiveJournal complied faithfully with the DMCA’s notice and takedown requirements but has nevertheless been exposed to expensive litigation and substantial, uncertain damages.
The potential red flags in this case were watermarks visible on the face of the uploaded photos. Some of the photos featured, in the words of the court, a "generic" watermark; others were marked with “Mavrixonline.com.” In the Ninth Circuit’s view, contra the district court, “[t]he existence of a watermark, and particularly a watermark with a company name, is relevant to the knowledge inquiry.” Applying the standard from Veoh/Shelter Capital, the question with red flag knowledge is not whether the watermarks are relevant to the knowledge inquiry but whether the presence of the watermarks on the photos should have made it "immediately apparent to a non-expert" that the photos were infringing. This strikes me as doubtful. The presence of a watermark on a photo can, but does not necessarily, convey that the person who placed the watermark did so for the purpose of asserting copyright in it. For example, in the market for stock photos, when an image advertised with a watermark is licensed, the licensor will often provide the licensee with a clean, unmarked version of the image file. In that context, the mark seems to be a tacit (though not explicit) assertion of copyright. In other circumstances, however, photos are posted with watermarks not as a tacit assertion of rights but as an indicator of origin, specifically which website was first to post the photo. This happens in the case of "spy" shots for new models of cars, which at least some photographers actually expect and want to be shared, albeit with the built-in attribution that comes with a watermark. Given the different messages that watermarks can convey and the absence in this case of a copyright symbol or the word "copyright" from the watermarks, it seems a reach to say that infringement would have been immediately apparent to a hypothetical non-expert moderator. It would have been easy enough for Mavrix to incorporate a copyright symbol or the word "copyright" into its watermarks. If it had done that, the message it sought to convey (i.e., that the images were protected by copyright and were not to be copied without permission) would have been more plausibly "immediately apparent."
By failing to apply the Veoh/Shelter Capital red flag standard more rigorously to the facts of this case, the Ninth Circuit has expanded the reach of red flag knowledge and thereby narrowed the scope of the safe harbors. Whether LiveJournal's individual moderators, assuming they were LiveJournal's agents, had actual knowledge that individual Marvrix photos were infringing is an appropriate question to put to a jury. It should not, however, be for the jury to decide red flag knowledge in this case, given the nonobviousness of the message that Marvrix's watermarks convey.