This blog post is excerpted from our filing in response to the U.S. Copyright Office's 2016 Notice and Request for Public Comment on notice and takedown practice under the Digital Millennium Copyright Act (DMCA). The entire filing is available here.
16. How effective is the counter-notification process for addressing false and mistaken assertions of infringement?
We have not seen studies or significant public data on this question, though there will be useful information in the study just published by Urban, et al.97 Based on our own experience and discussion with other practitioners, we believe that it is rare for users to file counter-notices. Counter-notices certainly appear to be far less common than the improper removals that they are intended to counteract.
A handful of companies track counter-notices in their transparency reports. These companies don’t appear to aggregate the data over time and, in some cases, they track it using non-parallel categories so that comparison is difficult. For example:
- Twitter’s transparency report includes counter-notices, though seemingly only for tweets, not for Vine or Periscope material. Twitter’s latest semiannual report says that 56,971 tweets were withheld pursuant to DMCA notice, and the company received counternotices for 65 or 0.11%, all of which led to the restoration of the targeted content.98
- Tumblr’s June 2015 report states that of the 77,357 posts that were removed pursuant to takedown notices, 0.08% were restored using the counter-notice process.99 Tumblr received additional counter-notices that it did not honor, though it is unclear whether those counter-notices were rejected based on substantive copyright problems or formal noncompliance with section 512(g).
- Github reports some counter-notices, but counts them in the same category with retracted DMCA notices, reporting 17 counter-notices or retractions out of 258 notices, which may each have identified numerous alleged infringements.100
- Automattic reports, for its latest data set, that “less than 0.6% of the DMCA notices we received were later subject to a counter-notice; of those cases, we’re aware of further action being taken by the original DMCA complainant only once.”101
These tiny percentages are dwarfed by the portion of dubious DMCA removal requests that researchers have identified. (See studies reported in Appendix B [ed note for blog: also here].) Even if the studies are off by an order of magnitude in their estimates, the number of potentially mistaken or malicious notices still vastly exceeds the number of counter-notices.102
Importantly, the companies issuing detailed transparency reports may be relatively unique among small intermediaries in their commitment to protecting users and offering them a chance to counter-notice. It is unclear whether the thousands of other companies that have registered DMCA agents with the Copyright Office assume similar costs and inconveniences to provide a viable counter-notice process.
The ineffectiveness of the DMCA counter-notice process may be attributable to a number of causes:
- Intermediaries do not have significant incentive to bother notifying a user when her content has been removed based on a DMCA notice. The section 512(g) counternotification process refers only to 512(c) hosting providers, not other OSPs (e.g., search engines), for starters. Even for hosts, the only reason to offer counter-notice is to avoid liability to a user for improperly removing that user’s lawful content. But few believe that hosts face any meaningful risk of such liability, regardless of section 512(g). “Wrongful removal” claims against intermediaries in US courts have consistently failed, based on contractual and other defenses.103
- It is not clear how many users actually receive notice that their content has been removed. Not all hosting services tie user-generated content to user email addresses or other contact information. For those that do, users may supply email addresses from temporary or old and unmonitored accounts. It is accordingly difficult to know how many users really find out when their posts have been removed.
- The counter-notice process is intimidating. Consent to jurisdiction, which is a required element of a counter-notice under section 512(g)(3)(D), is a meaningful legal concession, and is particularly problematic for users who do not reside in the United States. In addition, the 512 section (g)(3)(C) statement under penalty of perjury, while analogous to the similar requirement for the notifier, is likely to be far more intimidating to individual users responding without benefit of counsel. And the cost of error for a user if she is mistaken about her copyright defenses is much higher than the cost of error for a copyright owner who is mistaken about her claims.
Collectively, these factors constitute a meaningful deterrent to counter-notice. The point we make here is not that Congress lacked the intent or policy basis for establishing the detailed hurdles for counter-notifiers in section 512(g). The problem is that, because counter-notice has not been an effective corrective for wrongful notices, section 512(g) alone cannot adequately protect Internet users from having their legal speech removed. For that reason, the other procedural protections for users in section 512, such as form-of-notice requirements and declarations of good faith by copyright owners, play a more important role than Congress may have foreseen. Robust interpretations and enforcement of those protections by the courts and the Copyright Office are critical to maintain the DMCA’s carefully structured balance. A more detailed discussion of these other protections in is included above in response to Question 12.
In (weak) defense of section 512(g), the transparency and expectation of procedural fairness created by the counter-notice process may be acting as a deterrent for some bad faith removal requests. It is possible, however, that the value of counter-notice is far exceeded by the value of public transparency about particular removals, such as those posted through Lumen or noted by the OSP on the page from which content has been removed. This transparency allows the identification of erroneous DMCA notices to be crowd-sourced across interested individuals online. To our knowledge there are no public datasets that would allow us to test this hypothesis.
97 See Jennifer Urban, et al., Notice and Takedown in Everyday Practice (2016), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2755628.
98Twitter, Transparency Report, Copyright Notices (2015), https://transparency.twitter.com/copyrightnotices/2015/jul-dec.
99 Tumblr, Copyright and Trademark Transparency Report (2015), http://static.tumblr.com/zyubucd/0uWntp2iw/iptransparencyreport2015a_upd....
100 GitHub, Transparency Report (2014), https://github.com/blog/1987-github-s-2014-transparency-report.
101 Automattic, Intellectual Property (2015), https://transparency.automattic.com/intellectual-property/intellectualpr....
102 This calculation assumes that the rate of counter-notice for the data sets discussed in Appendix B is similar to the rates reported in the transparency data discussed above. We see no reason to expect otherwise.
103 See, e.g., Song Fi, Inc. v. Google, Inc., 2015 WL 3624335 (N.D. Cal. June 10, 2015); Lewis v. YouTube LLC, 2015 WL 9480614 (Cal. App. Ct. Dec. 28, 2015); Sikhs for Justice “SFJ,” Inc. v. Facebook, Inc., 2015 WL 7075696 (N.D. Cal. Nov. 13, 2015).