Attacking the Notice-and-Staydown Straw Man

Cross-posted from the Center for the Protection of Intellectual Property (CPIP) Blog.

Ever since the U.S. Copyright Office announced its study of the DMCA last December, the notice-and-staydown issue has become a particularly hot topic. Critics of notice-and-staydown have turned up the volume, repeating the same vague assertions about freedom, censorship, innovation, and creativity that routinely pop up whenever someone proposes practical solutions to curb online infringement. Worse still, many critics don’t even take the time to look at what proponents of notice-and-staydown are suggesting, choosing instead to knock down an extremist straw man that doesn’t reflect anyone’s view of how the internet should function. A closer look at what proponents of notice-and-staydown are actually proposing reveals that the two sides aren’t nearly as far apart as critics would have us believe. This is particularly true when it comes to the issue of how well notice-and-staydown would accommodate fair use.

For example, Joshua Lamel’s recent piece at The Huffington Post claims that “innovation and creativity are still under attack” by the “entertainment industry’s intense and well-financed lobbying campaign” pushing for notice-and-staydown. Lamel argues that the “content filtering proposed by advocates of a ‘notice and staydown’ system . . . would severely limit new and emerging forms of creativity.” And his parade of horribles is rather dramatic: “Parents can forget posting videos of their kids dancing to music and candidates would not be able to post campaign speeches because of the music that plays in the background. Remix culture and fan fiction would likely disappear from our creative discourse.” Scary stuff, if true. But Lamel fails to cite a single source showing that artists, creators, and other proponents of notice-and-staydown are asking for anything close to this.

Similarly, Elliot Harmon of the Electronic Frontier Foundation (EFF) argues that “a few powerful lobbyists” are pushing for notice-and-staydown such that “once a takedown notice goes uncontested, the platform should have to filter and block any future uploads of the same allegedly infringing content.” Harmon also assumes the worst: “Under the filter-everything approach, legitimate uses of works wouldn’t get the reasonable consideration they deserve,” and “computers would still not be able to consider a work’s fair use status.” Like Lamel, Harmon claims that “certain powerful content owners seek to brush aside the importance of fair use,” but he doesn’t actually mention what these supposed evildoers have to say about notice-and-staydown.

Harmon’s suggestion that the reliance on uncontested takedown notices gives inadequate consideration to fair use is particularly strange as it directly contradicts the position taken by the EFF. Back in October of 2007, copyright owners (including CBS and Fox) and service providers (including Myspace and Veoh) promulgated a list of Principles for User Generated Content Services. These Principles recommend that service providers should use fingerprinting technology to enact notice-and-staydown, with the general caveat that fair use should be accommodated. Two weeks later, the EFF published its own list of Fair Use Principles for User Generated Video Content suggesting in detail how notice-and-staydown should respect fair use.

The EFF’s Fair Use Principles include the following:

The use of “filtering” technology should not be used to automatically remove, prevent the uploading of, or block access to content unless the filtering mechanism is able to verify that the content has previously been removed pursuant to an undisputed DMCA takedown notice or that there are “three strikes” against it:

1. the video track matches the video track of a copyrighted work submitted by a content owner;
2. the audio track matches the audio track of that same copyrighted work; and
3. nearly the entirety (e.g., 90% or more) of the challenged content is comprised of a single copyrighted work (i.e., a “ratio test”).

If filtering technologies are not reliably able to establish these “three strikes,” further human review by the content owner should be required before content is taken down or blocked.

Though not explicitly endorsing notice-and-staydown, the EFF thinks it’s entirely consistent with fair use so long as (1) the content at issue has already been subject to one uncontested takedown notice, or (2) the content at issue is at least a 90% match to a copyrighted work. And the funny thing is that supporters of notice-and-staydown today are actually advocating for what the EFF recognized to be reasonable over eight years ago.

While Harmon never explicitly identifies the “powerful lobbyists” he accuses of wanting to trample on fair use, he does link to the Copyright Office’s recently-announced study of the DMCA and suggest that they can be found there. Reading through that announcement, I can only find three citations (in footnote 36) to people advocating for notice-and-staydown: (1) Professor Sean O’Connor of the University of Washington School of Law (and Senior Scholar at CPIP), (2) Paul Doda, Global Litigation Counsel at Elsevier, and (3) Maria Schneider, composer/conductor/producer. These three cites all point to testimonies given at the Section 512 of Title 17 hearing before the House Judiciary Committee in March of 2014, and they show that Harmon is attacking a straw man. In fact, all three of these advocates for notice-and-staydown seek a system that is entirely consistent with the EFF’s own Fair Use Principles.

Sean O’Connor seeks notice-and-staydown only for “reposted works,” that is, “ones that have already been taken down on notice” and that are “simply the original work reposted repeatedly by unauthorized persons.” His proposal only applies to works that “do not even purport to be transformative or non-infringing,” and he specifically excludes “mash-ups, remixes, covers, etc.” This not only comports with the EFF’s recommendations, it goes beyond them. Where the EFF would require either a previous uncontested notice or at least a 90% match, O’Connor thinks there should be both an uncontested notice and a 100% match.

The same is true for Paul Doda of Elsevier, who testifies that fingerprinting technology is “an appropriate and effective method to ensure that only copies that are complete or a substantially complete copy of a copyrighted work are prevented or removed by sites.” Doda explicitly notes that filtering is not suitable for “works that might require more detailed infringement analysis or ‘Fair Use’ analysis,” and he praises YouTube’s Content ID system “that can readily distinguish between complete copies of works and partial copies or clips.” Doda’s vision of notice-and-staydown is also more protective of fair use than the EFF’s Fair Use Principles. While the EFF suggests that a previously uncontested notice is sufficient, Doda instead only suggests that there be a substantial match.

Unlike O’Connor and Doda, Maria Schneider is not a lawyer. She’s instead a working musician, and her testimony reflects her own frustrations with the whack-a-mole problem under the DMCA’s current notice-and-takedown regime. As a solution, Schneider proposes that creators “should be able to prevent unauthorized uploading before infringement occurs,” and she points to YouTube’s Content ID as evidence that “it’s technically possible for companies to block unauthorized works.” While she doesn’t explicitly propose that there be a substantial match before content is filtered, Schneider gives the example of her “most recent album” being available “on numerous file sharing websites.” In other words, she’s concerned about verbatim copies of her works that aren’t possibly fair use, and nothing Schneider recommends contradicts the EFF’s own suggestions for accommodating fair use.

Lamel and Harmon paint a picture of powerful industry lobbyist boogeymen seeking an onerous system of notice-and-staydown that fails to adequately account for fair use, but neither produces any evidence to support their claims. Responses to the Copyright Office’s DMCA study are due on March 21st, and it will be interesting to see whether any of these supposed boogeymen really show up. There’s little doubt, though, that critics will continue attacking the notice-and-staydown straw man. And it’s really a shame, because advocates of notice-and-staydown are quite conscious of the importance of protecting fair use. This is easy to see, but first you have to look at what they’re really saying.

Notice-and-Staydown and Google Search: The Whack-A-Mole Problem Continues Unabated

Cross-posted from the Center for the Protection of Intellectual Property (CPIP) Blog.

After my last post discussing the necessity for notice-and-staydown to help copyright owners with the never-ending game of whack-a-mole under the DMCA, I was asked to clarify how this would work for Google Search in particular. The purpose of my post was to express the need for something better and the hope that fingerprinting technologies offer. But, admittedly, I did not do a good job of separating out how notice-and-staydown would work for hosting platforms as compared to search engines. I think the whack-a-mole problem with hosting sites is indeed different than with search engines, and while fingerprinting may work well for the former, it’s probably ill-suited for the latter.

It’s clear enough how fingerprinting technologies can be applied to hosting platforms, and the simple fact is that they are already being deployed. YouTube uses its own proprietary technology, Content ID, while other platforms, such as Facebook and SoundCloud, use Audible Magic. These technologies create digital fingerprints of content that are then compared to user-uploaded content. When there’s a match, the copyright owner can choose to either allow, track, mute, monetize, or block the uploaded content.

There isn’t a lot of publicly-available information about how accurate these fingerprinting technologies are or how widely copyright owners utilize them. We do know from Google’s Katherine Oyama, who testified to Congress in early 2014, that “more than 4,000 partners” used Content ID at the time and that copyright owners had “‘claimed’ more than 200 million videos on YouTube” with the technology. She also acknowledged that “Content ID is not perfect, sometimes mistakenly ascribing ownership to the wrong content and sometimes failing to detect a match in a video.” Despite these imperfections, the scale of which she didn’t spell out, YouTube continues to offer Content ID to copyright owners.

Oyama also indicated that Content ID does not “work for a service provider that offers information location tools (like search engines and social networks) but does not possess copies of all the audio and video files that it links to.” This scenario is clearly different. When a site hosts content uploaded by its users, it can easily match those uploads to the content it’s already fingerprinted. When a site links to content that’s hosted elsewhere, it may not be possible to analyze that content in the same way. For example, the linked-to site could simply prevent automated crawling. Of course, not all sites block such crawling, but more would probably start doing so if fingerprinting were used in this way.

For Google Search, notice-and-staydown would likely not depend upon fingerprinting technology. Instead, the changes would come from: (1) delisting rogue sites, (2) promoting legitimate content, (3) improving auto-complete, and (4) ceasing to link to the very links that have already been taken down. These suggestions are not anything new, but it’s clear that Google has not done all it can to make them effective. This is not to say that improvements haven’t been made, and Google is to be commended for the work that it’s done so far. But it can and should do more.

Sticking with the example of The Hateful Eight from my prior post, it’s easy to see how Google Search promotes piracy. Using a fresh installation of Chrome so as not to skew the results, I need only type “watch hat” into Google Search before its auto-complete first suggests I search for “watch hateful 8 online.” After following this suggestion, the first seven results are links to obviously-infringing copies of the film. The first link points to the film at the watchmovie.ms site. A quick glance at that site’s homepage shows that it offers numerous (if not only) films that are still in theaters, including Spectre, Star Wars: The Force Awakens, Creed, and The Hateful Eight. In other words, Google’s first search result from its first suggested search term points me to an illicit copy of the film on a site that’s obviously dedicated to infringement.

I’ve never heard of watchmovie.ms, so I checked its WHOIS data. The site was registered on October 14th of last year, and Google’s Transparency Report indicates that it started receiving takedown notices for it just a few days later. To date, Google has received 568 requests to remove 724 infringing links to watchmovie.ms, but its search engine dutifully continues to crawl and index “about 39,000 results” at the site. And, for reasons I simply cannot fathom, Google prefers to send me to that pirate site rather than point me to Google Play (or to any number of other sites) where I can pre-order the film and “watch hateful 8 online” legally.

Making matters worse, at the bottom of the first page of search results for “watch hateful 8 online,” Google links to four DMCA takedown notices it received from copyright owners that resulted in five links being removed from the first page of results. These four notices, in turn, contain a combined total of 499 illicit links to The Hateful Eight that Google has already taken down. This truly boggles the mind. Google takes down five infringing links from one page of search results, consistent with the DMCA, but then it links to those five links along with 494 more such links. And these linked-to links are even better for infringers, since they’ve been vetted by Google as being worthy of taking down.

As the producer of The Hateful Eight, Richard Gladstein, relayed to The Hollywood Reporter, Google told him that it is “not in a position to decide what is legal and what is illegal online.” This is a cop out. In other venues, Google contends that it’s doing a lot to fight piracy. It submitted comments to the U.S. Intellectual Property Enforcement Coordinator this past October explaining how “changes made to [its] algorithm have been highly effective in demoting sites receiving a high number of takedown notices.” This shows that it is indeed in a position to determine what is “illegal online” and to take action against pirate sites. But simply demoting these sites is not enough—they should be delisted altogether.

Everyone knows that The Pirate Bay is a pirate site, yet Google indexes “about 914,000 results” from just one of its domains. As of today, Google has received 187,540 requests to remove 3,628,242 links to that domain, yet Google doesn’t choose to delist the site from its results. Nor does it even appear to be demoting it. The top three search results for “thepiratebay hateful 8” are links to infringing copies of the film on The Pirate Bay. It’s clear that these links are infringing, yet Google makes copyright owners continue playing whack-a-mole for even the most obvious infringements.

This isn’t how the DMCA is supposed to work. Congress even envisioned this whack-a-mole scenario with search engines when it wrote the DMCA. The legislative history provides: “If, however, an Internet site is obviously pirate, then seeing it may be all that is needed for the service provider [i.e., search engine or other information location tool] to encounter a ‘red flag.’ A provider proceeding in the face of such a ‘red flag’ must do so without the benefit of a safe harbor.” The Pirate Bay is “obviously pirate,” and Google knows as much even without the 3.6 million takedown notices it’s received. It knows the same thing about lots of pirate sites, including the other domain names contained in its list of greatest hits.

Google could be doing more to help copyright owners with the whack-a-mole problem, but so far, it’s only taken a few baby steps. And when defending its refusal to delist obvious pirate sites, Google contends that it’s defending freedom of speech:

[W]hole-site removal sends the wrong message to other countries by favoring over-inclusive private censorship over the rule of law. If the U.S. embraces such an overbroad approach to address domestic law violations (e.g., copyright), it will embolden other countries to seek similar whole-site removal remedies for violations of their laws (e.g., insults to the king, dissident political speech). This would jeopardize free speech principles, emerging services, and the free flow of information online globally and in contexts far removed from copyright.

Delisting The Pirate Bay from Google Search isn’t about favoring “censorship over the rule of law.” It’s about Google favoring the rule of law over blatant criminal infringement and doing its part to be a good citizen in the digital economy where it plays no small role. The comparison of the conduct of criminal infringers to the speech of political dissidents rings hollow, and delisting the most obvious and egregious sites does not threaten free speech. Google already claims to demote pirate sites, yet that doesn’t “jeopardize free speech principles.” Neither will delisting them. And as long as Google consciously decides to index known pirate sites with its search engine, the whack-a-mole problem will only continue unabated for copyright owners.

Endless Whack-A-Mole: Why Notice-and-Staydown Just Makes Sense

Cross-posted from the Center for the Protection of Intellectual Property (CPIP) Blog.

Producer Richard Gladstein knows all about piracy. As he recently wrote for The Hollywood Reporter, his latest film, The Hateful Eight, was “viewed illegally in excess of 1.3 million times since its initial theatrical release on Christmas Day.” Gladstein is not shy about pointing fingers and naming names. He pins the blame, in no small part, on Google and (its subsidiary) YouTube—the “first and third most trafficked websites on the internet.” While acknowledging that fair use is important, Gladstein argues that it has become “an extremely useful tool for those looking to distract from or ignore the real copyright infringement issue: piracy.” His point is that it’s simply not fair use when someone uploads an entire copyrighted work to the internet, and claims that service providers can’t tell when something is infringing are disingenuous.

Gladstein questions why Google and YouTube pretend they are “unable to create and apply technical solutions to identify where illegal activity and copyright infringement are occurring and stop directing audiences toward them.” In his estimation, “Google and YouTube have the ability to create a vaccine that could eradicate the disease of content theft.” While Gladstein doesn’t mention the DMCA or its notice-and-takedown provisions specifically, I think what he has in mind is notice-and-staydown. That is, once a service provider is notified that the copyright owner has not authorized a given work to be uploaded to a given site, that service provider should not be able to maintain its safe harbor if it continues hosting or linking to the given work.

No small amount of ink has been spilled pointing out that the DMCA’s notice-and-takedown provisions have led to an endless game of whack-a-mole for copyright owners. Google’s own transparency report boasts how its search engine has received requests to take down over 63 million URLs in the past month alone. And it helpfully tells us that it’s received over 21 million such requests over the past four years for just one site: rapidgator.net. Google’s transparency doesn’t extend to how many times it’s been asked to remove the same work, nor does it tell us anything about takedown requests for YouTube. But there’s no reason to think those numbers aren’t equally as frustrating for copyright owners.

The question one should ask is why these numbers aren’t frustrating for Google and YouTube, as they have to deal with the deluge of notices. Apparently, they don’t mind at all. According to the testimony of Google’s Senior Copyright Policy Counsel, Katherine Oyama, the “DMCA’s shared responsibility approach works.” Oyama notes that Google has spent tens of millions of dollars creating the infrastructure necessary to efficiently respond to the increasing number of takedown notices it receives, but many (if not most) copyright owners don’t have those kinds of resources. For them, it’s daily battles of manually locating infringements across the entire internet and sending takedown notices. For Google, it’s mostly-automated responses to take down content that otherwise brings ad-based revenue.

These struggles hit individual authors and artists the hardest. As the U.S. Copyright Office noted in its recently-announced study of the DMCA, “[m]any smaller copyright owners . . . lack access to third-party services and sophisticated tools to monitor for infringing uses, which can be costly, and must instead rely on manual search and notification processes—an effort that has been likened to ‘trying to empty the ocean with a teaspoon.’” What makes the process so frustrating—and futile—is the fact that the same works get uploaded to the same platforms time and time again. And any time spent sending the same takedown notice to the same service provider is time that is not spent honing one’s craft and creating new works.

Gladstein is correct: Service providers like Google and YouTube could be doing more. And, somewhat ironically, doing more for copyright owners would actually mean that both sides end up doing less. The obvious solution to the whack-a-mole problem is notice-and-staydown—it just makes sense. There’s simply no reason why a copyright owner should have to keep telling a service provider the same thing over and over again.

Those who object to notice-and-staydown often point out that the DMCA process is susceptible to abuse. Indeed, there are some who send notices in bad faith, perhaps to silence unwanted criticism or commentary. But there’s no reason to think that such abuse is the rule and not the exception. Google’s own numbers show that it complied with 97% of notices in 2011 and 99% of notices in 2013. That’s still a potentially-significant amount of abuse from notice-senders, but it’s also certainly a ton of intentional abuse from infringers whose conduct generated the legitimate notices in the first place. And the vast majority of those infringers won’t get so much as a slap on the wrist.

Turning back to Gladstein’s theme, discussions about fair use or takedown abuse are beside the point. The simple fact is that garden-variety copyright infringement involves neither issue. As CPIP Senior Scholar Sean O’Connor testified to Congress, “for many artists and owners the majority of postings are simply straight-on non-transformative copies seeking to evade copyright.” It’s this simple piracy, where entire works are uploaded to the internet for all to take, that concerns copyright owners most. Gladstein cares about the 1.3 million illicit distributions and performances of The Hateful Eight that are obviously infringing, not the commentary of critics that would obviously be fair use. And takedown notices sent because of these illicit uploads are anything but abusive—the abusers are the infringers.

The technology to make notice-and-staydown work already exists. For example, Audible Magic and YouTube both have the technology to create digital fingerprints of copyrighted works. When users later upload these same works to the internet, the digital fingerprints can be matched so that the copyright owner can then control whether to allow, monetize, track, or block the upload altogether. This technology is a great start, but it’s only as good as its availability to copyright owners. The continued proliferation of infringing works on YouTube suggests that this technology isn’t being deployed properly. And Google has no comparable technology available for its search engine, leaving copyright owners with little choice but to continue playing endless whack-a-mole.

Fortunately, the tides have been turning, especially as the technology and content industries continue to merge. And strides are being made in the courts as well. For example, a Court of Appeal in Germany recently held that YouTube has the duty to both take down infringing content and to make sure that it stays down. A quick search of YouTube today shows that The Hateful Eight, which is still in theaters, is legitimately available for pre-order and is illicitly available to be streamed right now. One wonders why YouTube chooses to compete with itself, especially when it has the tool to prevent such unfair competition. Regardless, there is real hope that Gladstein’s call for a “vaccine that could eradicate the disease of content theft” will be just what the doctor ordered—and that “vaccine” is notice-and-staydown.

[Update: This post unintentionally generated confusion as to whether I think notice-and-staydown means that fingerprinting technologies should be used with search engines. I do not think that would work well. I explain how search engines could do more to help copyright owners with the whack-a-mole problem in this follow-up post.]

Join Us at the Copyright and Technology NYC 2016 Conference on January 19

Cross-posted from the Center for the Protection of Intellectual Property (CPIP) Blog.

Co-produced by GiantSteps, the Copyright Society, and Musonomics, the Copyright and Technology NYC 2016 Conference will be held at New York University’s Kimmel Center on Tuesday, January 19th. CPIP is a proud Media Sponsor of the event.

The conference program is available here, and registration is still open here.

Jacqueline Charlesworth, General Counsel and Associate Register of Copyrights at the U.S. Copyright Office, will be the keynote speaker. The timing is very fortuitous, as the Copyright Office just last week announced a new study to evaluate the effectiveness of the DMCA’s safe harbor provisions in Section 512 of the Copyright Act. Among the issues to be studied are the “costs and burdens of the notice-and-takedown process” and “how successfully section 512 addresses online infringement.” These very issues will be discussed at the conference.

The conference panels will discuss topics including live streaming, notice-and-staydown, copyright management information, safe harbor eligibility, collective licensing, and mass digitization. CPIP’s Executive Director Matthew Barblan will moderate the panel on safe harbor eligibility, and CPIP’s Assistant Director Devlin Hartline will be a panelist discussing notice-and-staydown.

We hope you will join us for an exciting and intellectually rewarding event!

How Patents Help Startups Grow, Innovate, and Succeed

The following post first appeared on the Center for the Protection of Intellectual Property (CPIP) blog, and it is reposted with permission here.

Many academic studies of the patent system focus on the negative, extrapolating from anecdotes about a few bad actors to make the case that our patent system is broken and to bolster cries for legislation weakening patent rights. Precious few studies focus on the countless honest and hardworking patent owners whose inventive labors benefit us all. But understanding how patents support inventive enterprises is a crucial part of the equation, especially at a time when Congress is considering legislation that would make it extremely difficult for startups and individual inventors to enforce their patent rights.

In a newly-published working paper, The Bright Side of Patents, CPIP Edison Fellow Deepak Hegde, along with co-authors Joan Farre-Mensa and Alexander Ljungqvist, take a look how patents help startups grow. They show that, contrary to the claims made by several academics and activists, startups are not victims of the patent system. On the contrary, patents help startups become more successful and innovative.

The study finds that “patent approvals help startups create jobs, grow their sales, innovate, and eventually succeed.” When a startup’s first patent application is approved, its employment growth increases by 36% and its sales growth increases by 51% on average over the next five years. First-patent approval also has a strong causal effect on a startup’s continued ability to innovate, increasing the number of subsequent patent grants by 49% and increasing the quality of those patents by 27%. In fact, a startup with first-patent approval is twice as likely to end up listed on a stock exchange—a common indication of success for a startup.

Negatively affecting startups are delays in the patent application process and ultimate application rejections. For every year an ultimately-approved patent application is delayed, the startup’s employment growth decreases by 21% and its sales growth decreases by 28% on average over the following five years. Furthermore, each year a patent application is delayed, the average number of subsequent patents granted decreases by 14% while the quality of those patents decreases by 7%. And for each year of delay, the probability that a startup will go public is cut in half.

One big reason why patents help startups is that they make it easier to access capital from external investors. The authors find that patents serve to mitigate frictions in information between potential investors and startups. Patents play an important role by alleviating startups’ concerns about having their inventions misappropriated by investors and by alleviating investors’ concerns about the credibility, quality, and monetary future of the startups. Having access to capital in turn sets startups on a path of growth where they can turn ideas into products and services, generate jobs, increase revenue, and undertake further innovation.

What makes this study unique is its unprecedented access to the USPTO’s internal databases, which allowed the authors to evaluate detailed review histories of both approved and rejected patent applications. Prior studies only focused on approved applications, thus making it impossible to accurately separate out the economic and innovative effects. The authors here are able to demonstrate the direct benefits of patent protection with causal evidence from a large-sized sample—45,819 first-time patent applications filed by startups.

There is a surprising amount of criticism of the patent system today. Some claim that patents are a waste of time and resources for startups, useful only for defensive purposes. Others claim that patents actually harm startups. The authors here show that startups that secure patent protection are in fact more likely to succeed. As Congress considers yet another round of large-scale patent legislation, lawmakers need to understand the role that enforceable patent rights play in enabling startups to grow and succeed. This study is a great step in adding some much needed clarity to the ongoing patent policy debates.