Bittersweet DMCA Safe Harbor Defense Win in Ninth Circuit–Ventura v. Motherless (Catch-Up Post)
Motherless runs a UGC site for adult content. None of its content is licensed from content producers. It is primarily ad-supported (85%), with the remaining revenues coming from subscriptions (but only 0.2% of active users are subscribers) and sales of schwag. For a while, Motherless paid members to upload content (in schwag and later in cash); one bulk uploader of 300k files earned a paltry $200.
Motherless had a more laissez-faire content standard than most sites. Here is their masthead:
Motherless banned child porn, bestiality, and infringing files, but tolerated other UGC. (The site’s motto: “anything legal stays”). Motherless provided copyright owners with the technical ability to remove files they thought were infringing (without sending takedown notices). Motherless’ principal, Lange, and a contractor pre-screened all user uploads. For video, Lange used software that created 5 screenshots (at the 20%, 40%, 60%, 80% and 100% points in the video) and then Lange/the contractor reviewed the images for “obvious signs of child pornography, copyright notices, watermarks, and any other information that would indicate that the [material] contains illegal content or violates” the TOU. Lange reviews 80 screenshots a minute (i.e., less than 1 second per screenshot) and 30,000-40,000 images a day.
The plaintiff identified 33 clips that allegedly infringed its copyright. Some of the clips contained watermarks placed by third party aggregator sites, but not from the plaintiff itself. The plaintiff never sent takedown notices for those clips. Nevertheless, the plaintiff sued for copyright infringement. The district court granted summary judgment in favor of Motherless. Ventura Content, the copyright owner, appealed. The Ninth Circuit, with one judge dissenting, affirmed the dismissal.
Material Stored at the Direction of the User: The first question was whether the infringing material was stored at the direction of the user.
Ventura argued that the owner of Motherless initially uploaded hundreds of thousands of files to seed the site, but the court says this is irrelevant because Ventura didn’t claim this initial batch contained any of its materials.
Ventura also argued that the content was not stored at the direction of users because Motherless’ owner scanned thumbnails of the content before posting. The court concludes that Motherless is not engaged in any editorial control. Unlike Mavrix, where the Ninth Circuit said that editorial decisions on what to post could create a factual question about whether content was posted at the users’ direction, this case only involves screening out of objectionable material. The court also concludes that neither tagging, group, nor algorithmic suggestions for “most popular” material change the user-submitted status of the material.
The court also says that storage at the direction of users does not only apply to “electronic storage lockers”—materials that are viewable by other users also qualify.
Actual/Apparent Knowledge of Infringing Content: Ventura argued that Motherless’s human pre-review of thumbnails resulted in knowledge of infringing content or facts from which infringing activity should have been apparent. Ventura says watermarks in the content triggered actual knowledge, but the court says the watermarks did not signal that Ventura owned the content in question.
Ventura argued also that the high production value of the content should have been a tipoff. But the court says this too is not legally relevant. Professionals often post content online to attract attention, and similarly, amateurs often produce professional quality content. In any event, the line between amateur and professional is becoming increasingly grey. The court also takes a dig at Ventura’s content and says nothing distinguishes its content from good quality amateur creations. Ultimately, nothing points to actual knowledge on Motherless’s part.
Ventura relied on similar facts to support so-called “red-flag” knowledge. The court says that is important is not that there is generally knowledge of infringing activity, but facts pointing to infringement of the content at issue. Citing to UMG v. Shelter Capital and the Second Circuit’s opinion in Vimeo, the court says that the bar for apparent awareness is high:
The copyright owner must show knowledge, actual or red flag, for the videos that infringed its copyright and are the subject of its claim. And for red flag knowledge, infringement must be apparent, not merely suspicious. Congress used the word “apparent,” not “suspicious” or some equivalent. Ventura, not Lange, is in charge of policing Motherless for its copyrighted material. Congress could have put the burden of policing infringement in suspicious circumstances on the provider, but it instead put it on the copyright holder.
The court distinguished Motherless from the site in Fung, where the site marketed itself as a pirate site, the site’s operators solicited copyrighted material, and it assisted users in viewing such material. Since Motherless had no actual or apparent knowledge of Ventura’s copyrighted material, it had apparent knowledge. (The complaint triggered an exchange between the parties where Ventura provided Motherless URLs of infringing material, and Motherless promptly removed or disabled the material.)
Right and Ability to Control Infringing Activity (and Direct Financial Benefit): The court quickly breezes through this requirement, finding that while Motherless could remove content and enforce its guidelines, it did not exert “substantial influence” over the activities of users. It did not direct users to upload specific types of content or curate the content users did upload. Its incentive program was negligible, in the court’s view. Finally, it was advertising-driven and did not receive any financial benefit directly attributable to the infringing activity.
Repeat Infringer Policy: The court says that Motherless took sufficient steps to terminate repeat infringers. It did not have a detailed policy, it did not inform its contractors of any standards they must use, it did not keep a log of previous terminations, and did not have any automated means to prevent previously terminated members from re-joining the site. Still, in the majority’s view, Motherless was a one-man shop, and as such, he could be excused from not having detailed written procedures regarding repeat infringer termination. Despite his testimony being inconsistent on how exactly he went about terminating repeat infringers, the court says that there is little evidence of “proven failures to terminate.” According to the court, the evidence showed that Motherless terminated between 1,320 and 1,980 users and only nine alleged repeated infringers had slipped through.
A dissenting Judge Rawlinson disagreed that the evidence warranted summary judgment in favor of Motherless on this issue. He saw a “gargantuan” issue of fact. He voiced serious disagreement on the propriety of summary judgment based on sufficient disputes of fact regarding the sufficiency of Motherless’s repeat infringer policy.
This is a helpful ruling for platforms in several respects.
Both large and small sites manually review content, and the ruling clearly emphatically states that manual review and the use of discretion in excluding content does not make the content attributable to the site owner, rather than the uploader. A contrary ruling would result in a disincentive to manually review content. While the DMCA and Section 230 are different ways a service provider can be shielded, it’s good to see that reviewing content for compliance with a set of acceptable use standards will not necessarily endanger protection under either rule. (As I read the opinion’s discussion of the manual review process, it was still unclear what obligation a site had to expunge content that appears to be infringing.)
Most importantly, it endorses a view of red-flag knowledge that is fairly narrow. Cases are few and far between articulating exactly what is sufficient for red-flag knowledge and the standard remains annoyingly vague. Nevertheless, cases such as this one that say red-flag knowledge only arises in the extreme circumstances are undoubtedly helpful.
The dispute on the panel centers around how formal a repeat infringer policy needs to be. The majority takes a flexible approach, despite mixed evidence on the issue. (On a vaguely related note, Eric has flagged how the copyright office DMCA renewal requirement could unnecessarily penalize less sophisticated website.)
Ventura has moved for rehearing en banc. Its petition is pending. It will be interesting to see if enough judges vote to rehear the case.
I loathe blogging DMCA safe harbor opinions. All too often, they are long, ponderous, and irresolute, and they take me forever to blog. I had always planned to blog the district court opinion in this case but it was too long and detailed to complete. Then, this opinion has festered in my blogging queue for weeks (Venkat completed his first draft days after the opinion; all subsequent delays are due to me). I might need to admit that 50+ page-long DMCA safe harbor cases may be beyond my blogging capacity.
I see this opinion as a good news/bad news opinion, so I’ll divide my remarks that way:
* Citing Perfect 10 v. ccBill, the court says flatly: “The Digital Millennium Copyright Act places the burden of policing copyright infringement on the copyright owner.” The court provides a pretty good summary of how the DMCA should work:
A service provider must delete or disable access to known or apparent infringing material, as well as material for which he receives a statutorily compliant takedown notice. He must also terminate repeat infringers when appropriate. The copyright owner, not the service provider, has the burden of policing infringement. But the service provider, to maintain its shield, must respond expeditiously and effectively to the policing. If these conditions are met, the service provider will not be financially liable for infringing material on his website. The details, of course, get complicated, and we must address those complications.
* screening out illegal material doesn’t change the remaining material as still being stored at users’ direction. “We find it counterintuitive, to put it mildly, to imagine that Congress intended to deprive a website of the safe harbor because it screened out child pornography and bestiality rather than displaying it. Instead, we read section 512(m) to say that Congress expressly provided that such screening does not deprive a website of safe harbor protection.”
* the court denigrates the “red flags” nomenclature, saying “the sports metaphor is no more helpful than the statutory word.” Instead, the court properly focuses the analysis on whether infringement was “apparent” (the statutory language), and “infringement must be apparent, not merely suspicious.” Citing UMG and Io, the court says “we look at objective facts and circumstances from which the specific infringement would be obvious to a reasonable person.” Citing UMG and Vimeo, the court reiterates that generalized knowledge of infringement isn’t enough: “It is hard to imagine that a site with 12.6 million pictures and video clips uploaded by users would not contain some material that users had uploaded without authorization,” but that’s not enough to constitute apparent infringement.
* the court rejected the bogus argument that Motherless should have known the clips were infringing because of their production values. “An ordinary person who had not studied movie-making and personally made movies would likely be oblivious to the professionalism that the expert report identifies. This is not to criticize the quality of Ventura’s videos. The apparent amateurism may be a skilled professional means of giving them an appearance of authenticity. But nothing about their professionalism would inform Motherless that they were infringing or would make infringement apparent.” Thus, the court says:
even if it were obvious to a reasonable person that some of the material on the site must be infringing, that is not enough to lose the safe harbor. It must be obvious that the particular material that is the subject of the claim is infringing. Here, it would not be obvious to a reasonable person that the clips excerpted from Ventura movies were infringing.
* Service providers don’t need to terminate every repeat infringer in order to qualify for a “reasonable” implementation of a repeat infringer policy. The court says: “Eligibility for the safe harbor is not lost just because some repeat infringers may have slipped through the provider’s net for screening them out and terminating their access. ”
It’s also favorable that the court says a repeat termination policy doesn’t need to be written down and can be applied flexibly, at least in a small operation where a single decision-maker handles all matters. I’m not sure, however, if any sites larger than a single decision-maker could get the same result.
* Motherless won at summary judgment instead of trial, even though the dissent would have required a trial.
* the court’s jujitsu of the Mavrix v. LiveJournal language about files stored at users’ direction is problematic: “Lange does not exercise judgment in what to host. His editing is limited to the kind protected by section 512(m), screening out illegal material….[there is no evidence] Lange actually exercises judgment about what to host beyond his screening out child pornography, bestiality, and infringing material.”
So the court tap-dances around a key topic: if the site had exercised “editorial control” over user content, would it have been categorically denied the DMCA online safe harbor? The facts in this case were stylized because Motherless’ philosophy of “anything legal stays,” but that is one tail end of the spectrum. Most UGC sites take a more active approach of moderating legal but objectionable content. Has the 9th Circuit set up a standard that will foreclose DMCA safe harbor protection for those services. At minimum, this becomes another expensive point of litigation. Thus, this case could end up as a net strategic loss for most UGC sites, including YouTube.
* we know the legal standard for right/ability to control is “substantial influence,” but we still don’t have any clear idea what that means. The court simply runs through a bunch of facts to distinguish the case from Fung:
Nothing in the record suggests that Motherless told its users what to upload. Its homepage welcomed users to “a moral free zone where anything legal stays.” It did not curate uploaded content in any meaningful way, nor did it reject unpopular groups or content. Motherless deleted only user-created groups that contained little or no content, and it started deleting bestiality content due to legality issues raised by European advertisers.
So, if any of these facts change, does that mean the service provider does exert “substantial influence” over users? Notice again how Motherless’ limited content moderation efforts helped it here, though most sites wouldn’t be in the same position.
* the court has some terrible dicta about what might constitute a “standard technical measure”: “One can imagine a digital version of the old c in a circle (©) automatically triggering the uploading software to exclude material so marked by the copyright owner. ” Oh god, no.
* this litigation has been ongoing for 7 years against what is basically a one-person shop. If the business hadn’t involved porn, the defendant almost certainly would have run out of money long ago. So the courtroom result may be good, but like the Veoh case, it may be too much of a Pyrrhic victory for most defendants.
And to compound the offense, Ventura never sent Motherless 512(c)(3) takedown notices. Ventura apparently preferred to spend years pummeling a defendant in court rather than just resolving the problem by sending takedown notices as Congress contemplated. SIGH.
As a final kicker, the court denies Motherless its attorneys fees, even though Ventura could have sent 512(c)(3) notices and saved both litigants what is likely hundreds of thousands of dollars of litigation costs. As with Veoh, an attorneys’ fee award won’t resurrect dead defendants, but the absence of a fee award can reinforce the injustices inflicted by plaintiffs.
Case citation: Ventura Content, Ltd. v. Motherless, Inc., 2018 WL 1310182 (9th Cir. Mar. 14, 2018)
Bonus Track (from Eric): Bad things are brewing in the ALS Scan v. CloudFlare lawsuit in the Central District of California. By now, you already know ALS Scan, a pornography company that has been bringing Internet copyright infringement cases for 2 decades. CloudFlare, as you also know, acts as a content delivery network for its customers. This both speeds delivery of files and allows CloudFlare to protect sites from DDOS attacks. CloudFlare’s unique status in the Internet stack appears to flummox the court.
CloudFlare requires copyright takedown notices to be submitted via its webform, in which case CloudFlare passes along the notices to the customers without taking further action. In this case, ALS Scan sent emails notifying CloudFlare of specific customer URLs that allegedly infringed, and CloudFlare allegedly took no action on the email notices.
On March 13, the court (mostly) denied summary judgment to both sides, sending the case to trial. The court spent a surprising amount of time trying to define the test for contributory copyright infringement. Citing Perfect 10 v. Amazon, the court uses this test: the defendant “(1) has actual knowledge of specific infringing activity, (2) is able, but fails to take simple measures to stop that infringement, and (3) continues to provide services to known infringers.”
The court says ALS Scan’s emails are sufficient to raise a triable issue on knowledge. The court rejects CloudFlare’s claim that there is no tertiary liability. The court says CloudFlare’s caching services qualify as “substantial assistance,” which the court seems to convert into but-for causation. So the court says CloudFlare lacks volition for direct infringement, but nevertheless the court cites CloudFlare’s marketing against it: “Cloudflare, by its own estimate seeks to provide its customers, with a faster, safer, and more efficient means by which they can deliver their content, i.e. to infringe.” Finally, the court says CloudFlare has a simple method of stopping infringement: it can terminate “serial infringers.” Does this essentially create a rule where a single act of infringement committed by a customer’s user means CloudFlare must terminate its services across its entire customer network…?
In the same case, in February the court denied summary judgment to ALS Scan against defendant Steadfast. Steadfast apparently acts like a server farm, providing hosting services to imagebam.com, which in turn provided hosting services to users who allegedly infringed ALS Scan’s content. The court says “under Louis Vuitton and Fonovisa, Steadfast’s decision to continue to lease server space knowing that space is being used for infringement potentially gives rise to liability.” Furthermore, the court “Steadfast did something more than lease physical server space, including providing ‘basic management services’ to Flixya, ‘configuring’ the servers, and physically ‘maintaining’ the servers.” (What is the difference between “leasing physical service space” and configuring and maintaining those servers?). The court nevertheless denies SJ to ALS Scan because “Steadfast did not offer much in the way of services beyond the leasing of physical servers” (again, what does “leasing” entail?) and didn’t exercise as much control as the defendants in Fonovisa or Akanoc, and the parties dispute what Steadfast did and could do when notified of its customer’s alleged infringements. These fact disputes are also enough to deny ALS Scan SJ on Steadfast’s DMCA safe harbor.
In March, the court granted summary judgment to Steadfast. Though the court’s ruling is filled with plaintiff-favorable dicta, and the court rejects most of the defense arguments, the dispositive point is that Steadfast successfully got its customer, Flixya, to expeditiously remove infringing files following ALS Scan’s notice. The court says: “once those infringing acts actually occurred, i.e., the mole emerged, Steadfast took simple measures to cause the images to be removed, i.e. to successfully whack the mole.”
While the news is good for Steadfast, I guess, the facts of this case and the jumbled 9th Circuit precedent is clearly overwhelming this judge. I bet the other defendants will be relieved when they get past this judge and, if necessary, take it up with the 9th Circuit.