When Does Content Moderation Trigger Direct Copyright Infringement?–LoopNet v. CREXi (Catch-up Post)
[My standard lament about Westlaw’s delays processing opinions from the Central District of California.]
CoStar (LoopNet) and CREXi are involved in a 5+ year litigation war that has already reached the Ninth Circuit at least once. The docket has 1,200+ entries, so this case has paid for a lot of lawyers’ kids’ private school tuitions. (By my count, 22 different Latham & Watkins lawyers have made an appearance in the case for CoStar; 14 lawyers, mostly from Keker Van Nest, have appeared for CREXi). I’m going to focus on a copyright ruling from late June that, like seemingly all other aspects of this case, was litigated scorched-earth.
CoStar runs LoopNet, a “digital marketplace platform” for commercial real estate listings. CREXi runs a competitive platform. CoStar hires photographers to take photos of the subject properties. Allegedly, 40k+ of those photos have also been posted to CREXi. The court rejects summary judgment for some aspects of CoStar’s copyright infringement claims, queueing this case up for a very expensive trial.
Direct Copyright Infringement
CREXi argued that users uploaded the photos in question and therefore it did not have volitional conduct sufficient to support direct copyright infringement.
Defining what constitutes volitional conduct, the court says “the critical issues here are whether CREXi’s users ‘selected’ every image at issue and whether CREXi exercised any control over which images were displayed ‘beyond the general operation of its website.'”
Users can upload photos individually or via an API. The court says users select the photos in both circumstances. CoStar argued “but the algorithms,” but the court says the algorithms only boosted listings, not individual photos. CoStar also pointed to CREXi’s “e-blasts,” but CoStar didn’t show that any of its copyrighted photos appeared in those.
CREXi exercises quality control over the uploaded photos, and the court sends that issue to trial to evaluate:
whether, to the extent CREXi notified users of any issues, it required the user to change an image before the user’s listing could be displayed, or merely alerted the user to the issue and left the ultimate decision up to the user. If CREXi conditioned the display of user listings upon the use of images that satisfied CREXi’s quality control criteria, this would indicate CREXi exerted some control over whether and when photos were posted. If, however, users were free to disregard CREXi’s suggestions, CREXi’s conduct was not volitional because the photos were “of a type and kind selected by the [user] and at a time initiated by the [user].” Further, to the extent CREXi prevented users from using photos because there was some indication that others may have had a copyright over the photo, that conduct cannot support “volition.”
This indicates that some types of content moderation turn the publication of users’ uploads into CREXi’s volitional conduct, thus qualifying as direct infringement. I remain clear on which types of content moderation create those risks. This issue is sent to trial.
CREXi also builds listings for its users, either directly or through BPOs:
There is ample evidence that CREXi and its BPOs copied listing information, including images, from LoopNet when a listing could only be found on LoopNet, and that CREXi would take screenshots of photos or otherwise crop out CoStar’s watermarks from photos to build out listings on its website…The evidence shows that when building out listings for its users, CREXi and its BPOs were involved to some extent in selecting the images used and endeavored to fix issues with and avoid blurry images….
the process by which CREXi and its BPOs built listings was not automated—on the contrary, the process was very manual and depended on human input of data and images
In these circumstances, the court says “CREXi employees and its BPOs operate the copying system.” CREXi can be liable for the BPOs’ efforts because:
CREXi retained the BPOs to build listings, perform quality control, and make material available on CREXi’s website based on CREXi’s specifications, or that CREXi supervised the BPO’s work…. CREXi instructed its BPOs on how to build listings, including to remove the watermark from any CoStar-watermarked images before using an image in a listing
Thus, “CoStar has stated a prima facie case for copyright infringement as to the User-Directed images displayed before the lawsuit was filed.”
DMCA 1202
The 1202 issue gets sent to trial because:
CoStar’s use of the logo (and what information the logo actually conveys) has changed over time. CREXi provides evidence suggesting CoStar used the logo on photos it did not own.
…and…
Some evidence indicates that CREXi considered CoStar’s logo to be branding, not CMI, and treated it as such….But other evidence suggests that CREXi knew removing the logos from images before using on CREXi’s website might raise legal copyright issues…Evidence further indicates that before this litigation, CREXi instructed its employees and BPOs that, when photos were only available on CoStar, they should screenshot the photos and crop them to remove the watermark
DMCA 512 Safe Harbor
Repeat Infringer Policy
CoStar never sent any takedown notice to CREXi before suing, so that kind of ends the inquiry about CREXi’s repeat infringer policy. There’s more discussion in the opinion about the standards for repeat infringer policies. In particular, CoStar took the position that it’s unreasonable to issue a single strike to a user who has allegedly infringed multiple photos in the same listing. CREXi argued that it would be unfair to kick those users off without giving them a chance at rehabilitation, and the court said that was reasonable. (CREXi has a 6-strike policy for termination, which the court doesn’t expressly approve or reject).
However,
If CoStar’s watermark identified CoStar’s copyright ownership, CREXi arguably prevented CoStar from detecting infringement and sending takedown notices by cropping or having others crop the watermark out before images were copied onto CREXi’s website. This could also support CREXi’s red flag knowledge of infringement
This issue gets sent to trial.
Standard Technical Measures
[The fact that CoStar litigated the 512(i) issue is a good indicator of the scorched-earth nature of this lawsuit.]
There is evidence that CREXi advised its users to upload (or provide to CREXi to upload) un-watermarked versions of the photos at issue. However, CoStar cites to no authority or evidence demonstrating that copyright watermarks are “standard technical measures,” and the Court has identified no binding case law addressing the issue…the Court is “unable to determine on this record whether accessing websites is a standard technical measure”
So this issue gets sent to trial too.
Stored at User’s Direction
there are genuine disputes of fact regarding the nature and extent of CREXi’s post-upload communications with users as to User-Uploaded images. These same facts preclude summary judgment as to whether those images were stored at the direction of CREXi’s users. As for the User-Directed images, these were uploaded by CREXi and its BPOs, not by the users themselves. CREXi argues that it only uploaded images at the instruction of and with authorization from its users, and its “mechanical[ ] upload[ing]” of these photos constitutes “conduct narrowly directed towards enhancing the accessibility of” the photos. But the evidence indicates that (1) CREXi reviewed images for whether they were blurry, upside down, or contained watermarks, (2) CREXi cropped watermarks out of photos before listings were posted, (3) at least in some instances, CREXi exercised discretion regarding which photos to include in a listing, and (4) in some instances, CREXi may have built listings without permission from the brokers. Unlike cases where courts have found material was stored at the direction of a user, either CREXi or its BPOs—not its users—uploaded the User-Directed images, and the evidence reflects that they exercise at least some judgment in which images to include in CREXI-built listings. This conduct is not “narrowly directed” at enhancing the accessibility of content already on CREXi’s website, but instead done for the purpose of “making that infringing material accessible” on the site in the first instance.
Thus, the user-directed images aren’t covered by 512(c) at all, and the user-uploaded images and 512(c) gets kicked to trial.
Summary
The court rounds up its conclusions:
• Any instances of infringement that occurred before December 19, 2022 are time-barred.
• CoStar owns the images for which it seeks partial summary judgment.
• CREXi’s algorithm does not constitute volitional conduct.
• CREXi’s marketing blasts do not constitute volitional conduct.
• There is no evidence that CREXi refused to take down images—therefore, this cannot support volitional conduct.
• CREXi engaged in volitional conduct as to the UserDirected images displayed before this lawsuit was filed.
• CREXi’s BPOs were its agents and copied images based on CREXi’s instructions before this lawsuit was filed.
• Any images uploaded to CREXi’s website before June 19, 2017 cannot support CoStar’s DMCA claims.
• Safe harbor protection under 17 U.S.C. § 512(c) does not apply to the User-Directed images.
• CoStar’s “mass-scraping” of CREXi’s website cannot support an unclean hands defense because such conduct is unrelated to the claims in this case.
Despite its overall irresolute nature, CoStar got some key wins in this ruling. These rulings should translate into CoStar getting paid for at least some of the photos.
I imagine the court hoped that this ruling gave the parties enough information to settle before trial. Instead, as far as I can tell, this litigation machine is still churning out impressive quantities of billable hours.
Case Citation: CoStar Group, Inc. v. Commercial Real Estate Exchange, Inc., 2025 WL 2647269 (C.D. Cal. June 25, 2025)