'Google''s numbers on 'bogus' DMCA notices: a closer look

One of the great things about blogging is the ready availability of primary documents that can easily be linked to. Readers don't have to trust the blogger's characterizations of what someone else said; they can just click on the link to the article, document, court opinion, blog post, etc., and see if the original blogger accurately described the source material, or they can just read the whole thing to get additional detail. This phenomenon is also a good check on bloggers; at least in theory, they won't just write things that are at odds with what the primary documents actually say.

Well, at least that's the way it's supposed to work. But, unfortunately, not always. Here's what I mean: New Zealand is debating a proposed "graduated response" law that would force ISPs to terminate the accounts of repeat copyright infringers. As part of the debate, Google submitted comments to the Telecommunications Carriers' Forum, an association of New Zealand ISPs. Here's how Techdirt characterized the filing:

Some entertainment industry lawyers have been going around lately, pitching a fable that the DMCA isn't really that bad, since bogus takedown notices are somewhat rare. However, some new evidence from Google suggests quite a different story. Reader Slackr points us to some news about Google filing a comment on New Zealand's proposed new copyright law that would kick file sharers offline based on accusations rather than convictions.... While it's interesting alone that Google is participating in the process, even more interesting is what it has to say about its experience with DMCA takedown notices:

In its submission, Google notes that more than half (57%) of the takedown notices it has received under the US Digital Millennium Copyright Act 1998, were sent by business targeting competitors and over one third (37%) of notices were not valid copyright claims.

Google's point is that these types of laws are widely abused, and setting up such a system where punishment is handed out without any real due process is going to lead to an awful lot of mistakes. But, these stats are worth discussing just for what they say about the DMCA itself, and that myth that the process is rarely abused. From the numbers Google has seen, it's quite clear that the DMCA isn't just abused, it's regularly abused in ways that are both anti-competitive and chilling.

Sounds interesting. So what's the "new evidence from Google"? Well, Techdirt doesn't link to Google's filing. Instead it just links to an article from New Zealand PC World, which doesn't tell us much.

Curious, I took a few minutes to do some heavy-duty original research (i.e., a few Google searches), and guess what I found? That "new evidence from Google"? It's: 1) not "new"; 2) not "from Google"; and 3) doesn't remotely prove the point for which Techdirt (and Google) cites it.

Here's the actual Google filing. The so-called "new evidence" is actually a citation in footnote 3 (page 9) to a 2006 executive summary of an article by two copyleft academics, purporting to find problems with DMCA takedowns. See J. Urban & L. Quilter, Efficient Process or “Chilling Effects”? Takedown Notices Under Section 512 of the Digital Millennium Copyright Act. Full article here. So the "evidence" is not new -- it appeared in a 2006 article, and thus was collected well before that -- and it was not "from Google"; it was "from" academics with a clear point of view on copyright issues. (I should mention that I know and like Jennifer Urban, one of the authors, though we have very different views on these issues.)

So what about the actual claims that Techdirt relies on in its post? The first is that "more than half (57%) of the takedown notices [Google] has received under the US Digital Millennium Copyright Act 1998, were sent by business targeting competitors." But what the Urban/Quilter article actually says (and which Techdirt fails to note) is that the 57% refers only to requests for search engine result link removals. This has virtually nothing to do with entertainment companies sending notices about unauthorized copies of their works (which is what graduated response is all about). As the exec. summary itself says (p.10), "entertainment companies...choose not to send search engine complaints." Having been involved in the takedown process at a major entertainment company, I can confirm that is correct. (And the only time I ever saw one entertainment company ever send a takedown notice on a competitor's work was the rare instance of a clerical error or confusion about copyright ownership.)

The other claim that Techdirt cites is that "over one third (37%) of notices were not valid copyright claims." This misstates what the underlying report says. I'm not sure where the "37%" figure comes from; I can't find it in the executive summary that the Google filing cites (or in the full article). (The closest I could find was 30%. See p. 12 of the exec. summary.) Moreover, the report does not say that the relevant DMCA notices "were not valid copyright claims." Rather, it refers to claims about which the authors (who are quite unsympathetic to copyright owners and favor an expansive view of fair use) had "significant questions....where a genuine dispute related to copyright infringement or defenses would clearly arise." (pp. 11-12). There's a very...signficant difference between "not valid" and "significant questions [as determined by academics unsympathetic to copyright owners]."

And since Techdirt is focused on demonstrating that "entertainment industry lawyers" are telling "fables," let's examine the data in the Urban/Quilter article to see how many of the relevant Google notices were actually sent by these alleged fabulists. According to the full article (p. 658):

Out of all § 512(c) notices in either the Google or self-reported sets (303 in total), 10 relate to movies; 7 to music; 7 to games; 63 to software; 37 to photos; 8 to other graphics; 166 to text; 5 relate to whole websites; and 4 were undefinable.

That's right: the study only looked at 10 "hosting" takedowns to Google or other sites about movies, and 7 about music. As the article candidly says about the music notices (p. 659): "This is obviously far too small a sample to be statistically relevant, and almost certainly reflects selection bias relating to media coverage of the relevant subject matter." And how many of the movie notices were sent to Google? According to the article (p. 659): "Of the ten § 512(c) notices sent by companies in the film industry, only two were sent to Google." Two! One doesn't need to be an expert statistician to know that we shouldn't be drawing conclusions from such numbers. Much more about the selection bias and other potential problems with the Urban/Quilter study at Scrivener's Error.

Another important point: when "entertainment industry lawyers" (like yours truly) speak about DMCA notices to Google (and the relatively small number of bogus ones), they're usually referring to notices to YouTube (which is owned by Google). Google didn't acquire YouTube until late 2006 -- well after the Urban/Quilter study. The study, in fact, does not refer to YouTube at all. While I don't have exact numbers, I am confident that the vast majority of motion picture industry DMCA notices to Google over the past several years have in fact been notices to YouTube. (I don't know enough about DMCA notices to Google by the music industry to say how those numbers break down.)

The bottom line: the numbers cited by Google (and then repeated by Techdirt) say virtually nothing about the actual kind of notices most relevant to a "graduated response" program like New Zealand is considering: notices about movies and music. And the lesson: look at the underlying documents, not simply partisan (mis)characterizations of them.

(And for the umpteenth time: I am not saying that bogus DMCA takedown notices don't exist. They do.)

UPDATE: Techdirt's partial rowback here. And commenter "LostSailor"'s delightful fisking of the rowback here.