November 22, 2005
DMCA Online Safe Harbor Empirical Study by Urban and Quilter
By Eric Goldman
Jennifer Urban of USC and Laura Quilter have released a working report entitled "Efficient Process or “Chilling Effects”? Takedown Notices Under Section 512 of the Digital Millennium Copyright Act." To my knowledge, this is the first empirical study of the notice-and-takedown provisions of 17 USC 512, and it's a fascinating read.
17 USC 512 represents a compromise between content publishers and Internet companies brokered in 1998 as part of the much larger DMCA. Specifically, 512(c) and (d) provide an extra-judicial procedure for copyright owners to combat online infringements. These sections give online service providers ("OSPs") a statutory safe harbor for copyright liability due to users' activities. However, copyright owners can send a notice (a "512(c)(3) notice") to OSPs that users are infringing. In response, OSPs must either take down the specified material or lose the safe harbor.
Theoretically, there are several checks on the OSPs' decisions to remove content improperly. First, the statute (17 USC 512(g)) contemplates that users can provide their OSPs with a "counter-notification" that the users' content isn't, in fact, infringing. However, the statute doesn't require OSPs to respond or react to these counter-notifications, and 512(g) exculpation of liability is usually irrelevant based on the risk allocation provisions in the OSPs' EULA.
Second, the statute creates a new cause of action (512(f)) for sending bogus 512(c)(3) notices. 512(f) has been litigated infrequently. A Ninth Circuit case (the Rossi case) appeared to substantially weaken the cause of action, but the Online Policy Group v. Diebold and Marvel v. NCsoft cases indicate that the cause of action may have some teeth. Nevertheless, because 512(f) appears to have been litigated fairly infrequently, 512(f) acts as weak counterpressure on sending 512(c)(3) notices.
Third, market forces should punish OSPs who too quickly side with copyright owners at the expense of the OSP's customers. It's possible these market forces exist, but I've never seen any evidence suggesting or supporting that users choose OSPs based on copyright infringement takedown policies.
As a result, there is some concern that the extra-judicial adjudication procedure contemplated by 512(c) and (d) are skewed too far in favor of copyright owners. The fear is that copyright owners can lob in 512(c)(3) notices without much fear of negative consequences, even if the legal basis for those notices is questionable. Meanwhile, OSPs, when they receive questionable 512(c)(3) notices, have every incentive to take down content as requested (to preserve the 512 safe harbor) and no incentive to push back or resist the notice.
Urban and Quilter examine these issues by looking at a database of actual 512 notices. The data comes from the chillingeffects.org website, which acts as a repository of such notices and provides recipients of the notices with some information about what the notices mean. Unfortunately, this dataset is extremely skewed because the principal contributor of 512 notices is Google. While the data skew limits the report's ability to reach strong conclusions, the empirical data is revealing nonetheless.
The report's principal findings:
* 30% of 512 notices asserted copyright infringement where the notice raised "significant questions related to the underlying copyright claim, including fair use defenses, other substantive defenses, very thin copyright, or non-copyrightable subject matter"
* Many notices sought recourse for activity where the OSP would be covered by the more robust 512(a) safe harbor
* 9% of the notices significantly failed to comply with the requirements of 512(c)(3). [Eric's comment: I'm very surprised this number is so low; my own anecdotal experience is that very few people send 512(c)(3)-compliant notices]
* 57% of notices sent to Google sought to kick competitors out of Google's index
* 37% of notices sent to Google targeted International sites
Because of the skewed dataset, it's hard to make absolute policy inferences based on these findings. However, the findings may illustrate the structural deficiency of 512(c)(3) (a point Urban and Quilter also made in their presentation on this topic at Santa Clara University School of Law in October): the notice-and-takedown process was built before the rise of P2P file sharing and assumes that copyright owners needed expeditious response from a centralized host operator. The data also suggests that putative plaintiffs use 512 notices (or 512-style notices) for legal gripes far beyond the scope of copyright law.
The harder question is whether 512(c)(3) notices are leading to socially undesirable consequences, such as the shutdown of critics/parodists/whistleblowers. Any notice-based liability scheme will lead to mistaken judgments or overreactions by those on notice, so we should expect some curtailing of socially-desirable content from 512(c)(3). However, these costs may be offset by the benefit of giving copyright owners a potential means to quickly squelch harmful copyright infringement.
This study empirically reinforces that a lot of junk 512(c)(3) are zinging around cyberspace. Unfortunately, to resolve the broader policy point, we don't know how OSPs are responding to these notices. Are OSPs ignoring them? Are OSPs pushing back? Do OSPs instantly fold like a house of cards? Are OSPs making lots of mistaken/erroneous judgments? Without the OSPs' response, it's hard to know exactly how much damage is caused by junk 512(c)(3) notices.
Nevertheless, this report provides a very useful starting point to examine these questions. More broadly, I hope it prompts us to carefully examine why copyright owners get special extra-judicial leverage to address copyright infringement while virtually all other plaintiffs have zero recourse against OSPs under 47 USC 230, irrespective of the notices sent.
UPDATE 2: Scrivener's Error critiques the report.
I'm not too surprised that the numbers of deformity are so low, IN OUR SET. Notices submitted by an OSP have already been through some vetting. Specifically in the case of Google, they have an online set of instructions that's pretty explicit.
As to the quality and statutory compliance of notices as sent to all OSPs, I couldn't speculate, but it wouldn't surprise me if there were a much higher rate of statutory noncompliance.
Posted by: Laura Quilter at November 22, 2005 06:16 PM