Stanford Technology Law Review Symposium on Secondary IP Liability

By Eric Goldman

Last week we had a cyberlaw-fiesta in the Silicon Valley, the likes of which have been rarely (if ever) seen before. The Stanford Technology Law Review hosted a symposium on Thursday on Internet intermediary liability, then we hosted a 15 year retrospective on 47 USC 230 on Friday, then we had the inaugural Internet Law works-in-progress event on Saturday. I will do blog posts on all (the Internet Law works-in-progress post will be at Goldman’s Observations).

Today, my notes from the STLR event, Secondary and Intermediary Liability on the Internet. As usual, these are my edited notes/impressions and not verbatim transcriptions, so double-check before attributing anything to anyone. Further, I moderated the trademark panel, so I did the best to capture some high points while also doing crowd control over five strong personalities.



Julie Martin, Mozilla

Mozilla runs into UGC in two places:

1) Add-ons marketplace. 12,000 add-ons are available.

2) Personas gallery (browser skins). Users have posted 242,000 versions. Only 650 skins have been subject to takedowns = ¼%. Mozilla has never gotten a user putback request, even though some affected users have contacted Mozilla other ways. As a result, putting the burden of putbacks on users seems onerous to them.

Sometimes, copyright owners don’t give much thought to takedowns. When companies actually talk with Mozilla, they decide not to send takedowns because they reassess the possible impact of ticking off their fans. Skins are a great way to build brand loyalty, so why interfere with those folks’ relationship with the brand?

Mozilla struggles with jurisdictional questions. Also, when a takedown notice has 150 links, it’s only one letter for the copyright owner but imposes lots of research effort on Mozilla. Mozilla is also stressed by its exposure to high copyright damages.

Fred von Lohmann

Speaking for Google: Google receives takedowns covering 3M items each year in more than 70 languages. Google rejects a majority of takedown notices on receipt for deficiencies (not copyright, incomplete, etc), but even then it reaches out to the sender in each situation.

Not speaking for Google:

His paper will address rightsholders’ fears of a race to the bottom—that if a country grants immunity, intermediaries will do the minimum to comply, and that standard will become the benchmark globally.

Fred’s Q: do we see a race to the bottom empirically? Hypothesis: no, in fact we see the contrary. Ex: 35 hours of video uploaded to YouTube every minute, and Content ID automatically checks all of that. And racers to the bottom haven’t had much success in court.

So why do intermediaries do more than required? Is compliance due to service provider’s legal fears/uncertainty about DMCA? Compare 47 USC 230, where service providers have extremely robust immunity, but we don’t see a race to the bottom there. There are many business reasons why companies want licensed content.

[Eric’s comments: I like where Fred is going, but I wonder if the 230/512 comparison works. Many service providers do more than 230 “requires” because they are trying to establish their own reputation as a credible site—what I call the tertiary invisible hand in this paper. The same dynamic might play out for copyright, in that some service providers might want a reputation as a source of only authorized or licensed content. However, in many cases, a website can have a “legitimate” reputation even if it has a lot of infringing UGC; and there is significant commercial value in being known as a place where infringing content is available (while often there is little or no commercial value to being known as the website which harbors the kind of content that 230 protects). So it seems like a website’s motivations for policing against content immunized by 230 may not reach the same result with respect to infringing content.

However, I do agree with Fred that there are many reasons a website might license content even if not required under copyright law. I always make this point when I teach the Feist case—I ask the students if Feist might license Rural’s data even though the court said the data was completely unprotected by copyright law (the answer is yes, of course). Some reasons why a website might obtain a license even if the website isn’t worried about its copyright exposure:

• speed. The website might get content faster through a license.

• accuracy/authenticity. The website might want content versions that it knows haven’t been modified.

• comprehensiveness. The website might want everything in the content owner’s vault, not just what users have uploaded.

• reduced risk of harmful code. The website might fear that user-supplied versions of content have introduced malware or other risks to users.

• package deal. The website and content owner may have multiple points of business overlap that enable them to put together a “package deal.” This is why I still can’t believe Viacom and YouTube didn’t settle their disputes a long time ago.

• reputation. As I mentioned, a website might find it commercially valuable to advertise itself as a licensed source.

• morality. A website might decide that a license is the right thing to do.

I’m sure there are other reasons that haven’t occurred to me quickly. I trust Fred will taxonomize the motives in his paper.]

Peter Menell

We probably all agree that Sony should have been allowed to release VTR machines, but the Supreme Court got there with an overbroad rule. Grokster court needed to create “exception” to that rule. But the inducement rule focuses on mental state, not marketplace actions—not a good focus.

Some low-hanging fruit for doctrinal changes:

1) Statutory damage regime aren’t well-calibrated for the digital age. If we fixed this, the Viacom case might go away.

2) Strong sanctions for bogus takedowns.

Fred’s race-to-the-bottom isn’t the right Q. A few rogue actors can break the whole system.


Jacqueline Charlesworth: many content owners don’t have the resources to prepare and send takedown notices. Who is cheapest cost avoider? It’s the service.

[Eric’s note: Why, why, WHY do content owners keep repeating this trope? If Viacom cannot figure out which YouTube videos were posted by its own marketing people, there is simply no credible argument that YouTube is the lowest cost avoider to make an infringement determination ex ante. It *might* be true that after YouTube gets an infringement notice, then it is the cheapest cost avoider to prevent further harm. But that’s exactly what 512 codifies.]

Everyone can agree Grokster was bad. But content owners are grumbly about shift of value from content owners to tech providers.

Julie Martin: determining content legitimacy requires facts the service provider doesn’t have. When content owners reflect on it, many of them decide to leave up content—so there’s no way for a service provider to categorically decide what the copyright owner wants.

Matt Neco: in his experience, content owners wanted the service provider to license their content *just to set up the filter.* He also raised the concern that the “standard technical measures” requirements may, over time, favor incumbents who have made investments to accommodate those measures.

Fred von Lohmann: Content ID cost Google tens of millions of dollars to develop. But fingerprinting isn’t silver bullet. On a P2P network with encrypted traffic, fingerprinting is useless. YouTube’s existence spurs unprecedented creativity. The 106 rights are part of that, but all the defenses are an integral part too.

Mark Lemley: filters suck. They are both over- and underinclusive. Content owners can supplement Content ID with takedown notices. But no good solution for filters catching legitimate content. This puts burden on users to fix problems with the filters, and they aren’t in good position to do so.

Fred von Lohmann: Google worked hard to include recourse for user who gets filtered and easy to initiate file dispute. But content owners resent having to look at these disputes.


David Bernstein

The Tiffany v. eBay case’s result is inconsistent with Inwood, tort principles and Inwood’s progeny, but David thinks it reached the right result. eBay did a lot of things right, such as VeRO and filtering. David wants to avoid generalizing the ruling to create a notice-and-takedown system.

If a service provider can reasonably anticipate that counterfeiting is taking place, the service provider must take reasonable precautions. Notice-and-takedown isn’t enough—service providers should also filter for keywords, authenticate sellers and bar repeat offenders (and prevent whack-a-mole). We can’t say what reasonable precautions are; courts can decide. The courts should not focus on service provider state of mind but instead they should look to see if the problem actually got solved. Example: if someone is selling 5 Tiffany items, it’s probably not a casual seller of personal used good.

Mark McKenna

Secondary trademark liability rules don’t necessarily derive from the common law of torts. Traditionally, torts had three categories of secondary liability:

1) Vicarious infringement. Normally, we restrict the categories of vicarious liability pretty strictly.

2) Accomplice liability = liability for third party intentional tort based on requisite scienter. Look for particularized knowledge of forthcoming harm. Ex: a promoter’s promotion of an illegal boxing match that leads to a battery. This is similar to Grokster, although it focuses on more speculative knowledge.

3) Negligence = probabilistic harm of third party wrongdoing = defendant knows some chance of harm by other people. Ex: gun manufacturers who know some guns will be used criminally.

Typically, these categories involve situations where the third party harm also require scienter; not predicated on strict liability. IP is strict liability infringement, so it differs from typical tort principles.

Tiffany is trying to collapse these three categories.

We should ask what reasonable precautions eBay should take, so he agrees with David. But Mark may disagree with David over whether eBay took reasonable precautions. Tiffany’s standard may push to get eBay to block all infringing activity. Don’t forget costs of such zero-tolerance: it will also shut down of legitimate secondary markets. We should be talking more about proximate causation. When does third party intentional tortious conduct cut off upstream causation?

Stacey Dogan

Secondary trademark law is heading towards a “know it when we see it” standard. Courts are sorting between good guys and bad guys and deciding results accordingly. Good guys get the benefit of the doubt.

Tiffany v. eBay rule leaves room to go after bad guys. Why did district court spend so much time discussing eBay’s efforts? This helped support that eBay’s business model wasn’t designed to promote infringement; instead, it illustrated that eBay’s interests were aligned with TM owners’.

Court also noted willful blindness and inducement doctrines as tools to punish bad guys. Inducement in TM law = people whose goal is to help others infringe.

Copyright and trademark doctrines are different enough that we should be cautious importing copyright rules into TM law. In TM, there may be situations where TM owner and service providers interests are aligned, such as counterfeits. But in many cases, their interests diverge, such as trademark owners’ interests in shutting down secondary markets. Also, the range of legitimate TM secondary uses are even broader than range of legitimate secondary CR uses—any use that doesn’t confuse consumers is legitimate TM use.

Need to guard against the “freeriding” arguments. Appeals courts need to guide lower courts not to overreact to freeriding allegations.


Chris Sprigman: [Chris focused mostly on the Carroll Towing formula for reasonable precautions (Burden = Probability x Loss). He then hammered TM owners on the L:] The harm from counterfeits often overstated. Many people are looking for counterfeits because they can’t afford the original, but many of these people often end up buying the original later (his phrasing: “counterfeits are like a gateway drug”). Another faux harm: post-sale confusion. These situations may signal fashion “trends” that stimulate demand for the originals.

Mark McKenna: he thinks it’s right to think of Tiffany v eBay as a negligence case. It would have been nice if court had said so explicitly. We are still struggling with the theoretical split between TMs as consumer protection v. producer protection.

Mark Lemley: did Tiffany v. eBay set the floor of efforts required to police?

Mark McKenna: in tort framework, the reasonable precautions would adjust. But this isn’t clear because court didn’t talk in tort terms.