Google Inc.
25 Massachusetts Avenue, NW
9th Floor
Washington, DC 20001

Main 202 346.1100
Fax 202 346.1101
www.google.com

February 21, 2017
The Honorable Karyn Temple Claggett
Acting Register of Copyrights
U.S. Copyright Office
101 Independence Avenue, SE
Washington, DC 20559-6000
Re:

Section 512 Study: Request for Additional Comments
Docket No. 2015-7 (November 08, 2016)

Dear Acting Register Claggett:
Google Inc. (“Google”) appreciates the opportunity to submit additional comments in connection with the
U.S. Copyright Office (the “Office”) Request, Section 512 Study: Notice and Request for Public
Comment, 80 Fed. Reg. 251. We share the Office’s interest in ensuring the efficiency and effectiveness of
the safe harbor provisions contained in 17 U.S.C. § 512 for owners, online intermediaries, and users of
copyrighted works. Our initial comments (“Google’s April 1 Comments”) explained why the stable safe
harbor framework established by Section 512 is crucial not only to Google’s many online products and
services but to the growth of the Internet, and shared some of the steps we have taken to combat piracy
online. In these reply comments, we take the opportunity to explain the workings of Content ID, and to
address certain of the Office’s additional questions.
Introduction
Section 512’s notice-and-takedown regime has met Congress’s twin goals: it has “facilitate[d] the robust
development and world-wide expansion of electronic commerce, communications, research,
development, and education,” S. Rep. 105-190 at 1, while also providing copyright owners with a useful
tool to protect their works at scale without having to hire a lawyer or register works. 1 Moreover, Section
512 has produced a stable body of law2 on which those who invest in the next generation of online service
providers (“OSPs”) rely. 3 Section 512 has laid a foundation, atop which established OSPs have developed
innovative systems (like YouTube’s Content ID and other measures described in our April 1 Comments) 4
1

See Google’s April 1 Comments, at 1-4.

2

See id. at 5, 12-16.

3

See id. at 17-18.

4

Id. at 13.

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

that go far beyond what the DMCA requires and enable copyright owners to monetize uploads of their
content by third parties. At the same time, Section 512 preserves flexibility for less established platforms
without the resources to develop and implement such systems. 5
While the DMCA has not by itself eliminated rogue sites exploiting the copyrighted work of others, it has
succeeded in fostering collaboration and economic growth, and in driving many rogue actors from the
marketplace. The judicial record is clear, moreover, that rogue sites have found no shelter in the DMCA’s
safe harbors. Instead, this activity has successfully been driven out of the United States. The vast majority
of the remaining rogue sites have moved offshore, out of the reach of the DMCA; further tinkering with
the U.S. copyright regime is therefore unlikely to impact their behavior. As Ian Ballon put it at the San
Francisco Roundtable, the issue of rogue sites is not a DMCA safe harbor issue. 6 Online service providers
and rightsholders have nevertheless continued to collaborate to develop voluntary cross-industry efforts,
such as supply-side initiatives (making more lawful content available), coupled with follow-the-money
strategies, to address these ongoing challenges. 7 Early evidence suggests that these efforts are proving
successful. 8
In short, the DMCA has proven successful at fostering ongoing collaboration between rightsholders and
online service providers, a collaboration that continues to pay dividends both in the U.S. and in
international contexts.
Content ID
As described in our April 1 Comments, Content ID on YouTube is one example of the voluntary
measures Google has taken, above and beyond the requirements of the DMCA safe harbors, to collaborate
with and address the concerns of rightsholders. 9 As a voluntary measure, Content ID is not directly
relevant to the Office’s inquiry regarding Section 512. Nevertheless, because Content ID was mentioned
by several parties in the written and oral commentary presented to the Office in connection with the
Section 512 Study, it may be helpful to provide additional facts about the system and its use in practice.

5

See id. at 3-4.

6

Transcript of May 12, 2016 Section 512 Public Roundtable, at 263 (comments of Ian Ballon, Stanford Law Sch.
Ctr. for E-Commerce), available at https://www.copyright.gov/policy/section512/public-roundtable/transcript_0512-2016.pdf (“There’s a very big problem with rogue sites, but that’s a separate problem from the DMCA, and I
think that that’s important.”).
7

See Google’s April 1 Comments, at 7-8.

8

A Bunch of Weak Anti-Piracy Measures Are Still a Pest to Pirates, TorrentFreak (Jan. 7, 2017), available at
https://torrentfreak.com/a-bunch-of-weak-anti-piracy-measures-are-still-a-pest-to-pirates-170107/.
9

See How Google Fights Piracy, Google, available at
https://drive.google.com/file/d/0BwxyRPFduTN2TmpGajJ6TnRLaDA/view.

2

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

In order to use Content ID, rightsholders identify the works they own or control by providing reference
files of those works, along with metadata defining the scope of their copyright rights in those works.
Those reference files are fingerprinted such that even small portions can be identified even if major
changes are made to them (e.g., changes in dimensions, positioning, colors, or speed of playback). New
user videos uploaded to YouTube are compared against that database of fingerprints, and YouTube then
applies the business rules chosen by the relevant rightsholders. YouTube also scans the corpus of
previously uploaded videos in order to locate matches against new reference files provided by
rightsholders.
Using Content ID, rightsholders set various thresholds for declaring a match on their own content, as well
as what to do with matches. Rightsholders can decide to block content from ever appearing on YouTube
in the first place, to track usage of their material but otherwise take no action, or to monetize the claimed
content by placing advertising against it. Although the use of Content ID is not conditioned on licensing
any content for use on YouTube, more than 90% of all claims result in this third option, monetization. In
fact, the major record labels opt to monetize their content more than 95% of the time. To put it another
way, while Content ID today offers rightsholders the option to exercise “takedown-staydown” of their
works, the overwhelming majority of rightsholders prefer to use Content ID to “leave-up-and-get-paid.”
Content ID is currently used by more than 8,000 partners representing hundreds of thousands of
rightsholders. The system contains more than 50 million reference files, and has allowed rightsholders to
claim more than 400 million videos. YouTube has paid more than $2 billion to rightsholders from
Content ID alone, over and above other YouTube payments to rightsholders. After nearly a decade of
constant improvement, Content ID can now match on video, sound recording, and even the melodies of
musical compositions, each of which may be subject to different claims with different rules. 10
While Content ID is being used by thousands of partners, Content ID is not a tool that is appropriate for
everyone. It is an enterprise tool, providing a complex set of controls with platform-wide reach. Tools this
powerful and complex can have significant consequences when misapplied. For example, a local news
service that routinely claimed its news broadcasts forgot to exclude from its claim embedded NASA
footage relating to the Mars rover landing. This resulted in mistaken claims on many other stories using
the same, public-domain footage, and even a Content ID claim barring the video from appearing on
NASA’s own site. 11 For this reason, YouTube carefully evaluates applications from those who want to
become Content ID partners. In reviewing these applications, we take many factors into account,
including how popular the applicant’s works are on the platform, how many DMCA notices have been
10

We note also that many of the examples offered by other commenters concerning Content ID’s operation are years
old. As you might expect from any ongoing software project, the Content ID of 2007 or 2010 is barely relevant to
the performance of the current Content ID, which continues to evolve.

11

Timothy Lee, As Curiosity Touches Down on Mars, Video Is Taken Down from YouTube, Ars Technica (Aug. 6,
2012), available at https://arstechnica.com/tech-policy/2012/08/as-curiosity-touches-down-on-mars-video-is-takendown-from-youtube.

3

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

sent on their behalf, the quality of the DMCA notices previously sent, and whether a significant portion of
the applicant’s catalog is already represented by existing Content ID partners.
For rightsholders who are not good candidates to be Content ID partners directly, a growing number of
existing Content ID partners act as aggregators, representing smaller rightsholders. 12 Examples in the
music field include The Orchard Music, AdRev, CD Baby, Believe Music, Ingrooves, Kontor New Media
Music, Merlin, and Rumblefish. These are just a few of more than 30 such service providers operating on
YouTube, who collectively manage more than 13 million sound recordings on YouTube on behalf of
smaller rightsholders.
While Content ID has been a great success on YouTube, it is worth stressing again that its success is
closely tied to its development as a voluntary measure, suited to the particular context of YouTube. As we
explained in our April 1 Comments, the technical feasibility of such an automated system depends on the
nature of the online service and the nature of the works to be identified.13 What is technically feasible for
a public video hosting service like YouTube is not necessarily feasible for a private, encrypted messaging
or file lockering service, for a service that hosts non-video content, or for a service that primarily relies on
sharing links rather than hosting content. For this reason, “staydown” obligations on service providers
generally are better left to voluntary efforts, rather than legal mandates. In fact, Facebook, Tumblr,
Twitch, SoundCloud, DailyMotion, and Scribd have all described similar voluntary efforts in their
submissions to this Study.
The U.S. Copyright Office’s Additional Questions
Characteristics of the Current Internet Ecosystem
Question 1: As noted above, there is great diversity among the categories of content creators and
ISPs who comprise the Internet ecosystem. How should any improvements in the DMCA safe
harbor system account for these differences? For example, should any potential new measures, such
as filtering or stay-down, relate to the size of the ISP or volume of online material hosted by it? If
so, how? Should efforts to improve the accuracy of notices and counter-notices take into account
differences between individual senders and automated systems? If so, how?
Because the safe harbor system is working (as described in the Introduction), and because investors and
others rely on the stability it has provided, imposing mandatory additional obligations on service
providers, such as filtering, or stay-down, is unwarranted. As described above, many large OSPs have
voluntarily developed systems that go far beyond the floor set by Section 512. 14 But voluntary methods
12

For more information, see the YouTube Creator Services Directory, available at
https://servicesdirectory.withyoutube.com/.
13

Google’s April 1 Comments, at 10.

14

See, e.g., Comments of Facebook, Inc., at 4, 6-8, In re Section 512 Study: Notice and Request for Public

4

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

that work in one context (e.g., Content ID for YouTube) cannot be freely transported to another (e.g.,
search engines and social networks, where the content at issue is not uploaded or otherwise possessed by
the OSP and thus is unavailable for scrutiny). 15 The virtue of Section 512 is that it allows flexibility;
imposing stringent up-front requirements on large OSPs would displace this flexibility and produce little
in the way of tangible results beyond disruption of the settled expectations that have driven investment.
Moreover, we are not aware of any principled or technically sound basis on which to discriminate among
OSPs with respect to mandatory copyright obligations. As was pointed out during the roundtable sessions,
neither the size of a company nor the number of users it serves is a good proxy for whether copyright
infringement is a problem on a particular platform or whether an OSP can afford to shoulder additional
legal and technical obligations to protect the interests of copyright owners. Drawing arbitrary lines to
distinguish how copyright laws apply to different OSPs is likely only to encourage gamesmanship and
create market distortions that hinder economic growth and competition.
Question 2: Several commenters noted the importance of taking into account the perspectives and
interests of individual Internet users when considering any changes to the operation of the DMCA
safe harbors. Are there specific issues for which it is particularly important to consult with or take
into account the perspective of individual users and the general public? What are their interests,
and how should these interests be factored into the operation of section 512?
It is important to remember that in the digital age, users are also creators. The world is in the grip of the
biggest explosion of creativity in history. 16 People everywhere use new digital tools to create, and can
share their works worldwide through new platforms. In particular, it is easier than ever for creators large
and small to connect with an audience, build a fanbase, and earn a living without traditional
intermediaries. Every day, more than 1,000 YouTube channels cross the 1,000-subscriber milestone, and
in the last 12 months, YouTube has paid artists more than $1 billion from ad revenue alone. Many of
these are individual artists who are reaching their audiences directly and earning a living. Increasing
burdens on OSPs and decreasing legal certainty about the channels through which these new creators
make their living threatens this renaissance. Accordingly, if the Office intends to get a complete picture of
the functioning of the DMCA safe harbor system, it should reach out to these new creators on platforms
like YouTube, Instagram, the iTunes App Store, Google Play, Etsy, Twitch, and DeviantArt, to name just
a few.
Comment (No. 2015-7), available at https://www.regulations.gov/contentStreamer?documentId=COLC-2015-001390724&attachmentNumber=1&contentType=pdf; Comments of Soundcloud Operations, Inc., at 13-14, In re Section
512 Study: Notice and Request for Public Comment (No. 2015-7), available at
https://www.regulations.gov/contentStreamer?documentId=COLC-2015-001390151&attachmentNumber=1&contentType=pdf.
15

Google’s April 1 Comments, at 10.

16

See Michael Masnick & Michael Ho, The Sky Is Rising 2, Floor64 (Jan. 2013), available at
https://www.techdirt.com/skyisrising2; Michael Masnick & Michael Ho, The Sky Is Rising, Floor64 (Jan. 2012),
available at https://www.techdirt.com/skyisrising/.

5

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

Operation of the Current DMCA Safe Harbor System
Question 3: Participants expressed widely divergent views on the overall effectiveness of the DMCA
safe harbor system. How should the divergence in views be considered by policy makers? Is there a
neutral way to measure how effective the DMCA safe harbor regime has been in achieving
Congress’s twin goals of supporting the growth of the Internet while addressing the problem of
online piracy?
The best way to measure whether Congress’s twin goals have been met is to consider the state of the
online industry and the state of the content industries. Streaming services are displacing piracy. 17 Music
consumption is exploding worldwide 18 and traditional-music-industry revenues are growing. 19 The
Internet is thriving. Television is in a second Golden Age. 20 These trends show that the DMCA has been
effective in meeting both of Congress’s goals.
Those expressing negative views about the DMCA safe harbor often cite the large number of takedown
notices submitted to Google Search as evidence that the safe harbor is ineffective.21 That premise does not
follow—copyright owners would not submit a large volume of takedown requests unless they believed
doing so was worthwhile. And other explanations for the increase in notices are more persuasive. First, in
recent years, Google has streamlined its submission process, enabling rightsholders to send more notices
more easily (while still continuing to reduce the average time to resolution to under six hours). Second,
Google has provided new incentives to make heavy use of the DMCA takedown system. We now use the
number of valid DMCA requests a domain has received as one of the inputs in making ranking
determinations in search results, so rightsholders seeking to take advantage of this signal have further

17

Luis Aguiar & Joel Waldfogel, Streaming Reaches Flood Stage: Does Spotify Stimulate or Depress Music Sales?,
at 4 (European Joint Research Ctr., Working Paper 2015/05, 2015), available at
https://ec.europa.eu/jrc/sites/jrcsh/files/JRC96951.pdf (concluding that an increase of 47 streams has the effect of
one less pirate download); id. at 36 (presenting data and analysis to this effect).
18

“Music Consumption Exploding Worldwide” is in fact the subtitle of the IFPI’s own 2016 Global Music Report.
IFPI, Global Music Report (2016), available at http://www.ifpi.org/downloads/GMR2016.pdf.
19

Id. at 8 (noting that global music revenue grew by 3.2% in 2015).

20

See Jon Erlichman, The ‘Golden Age of TV’ Has a Lot of People Worried – Here’s Why, Fortune (Jan. 18, 2016),
available at http://fortune.com/2016/01/18/golden-age-tv-peak/ (noting a 94% increase in the number of scripted
shows being created, and considering whether there may be negative market effects from too much television).
21

See, e.g., Transcript of May 2, 2016 Section 512 Public Roundtable at 54, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-02-2016.pdf (comments of Victoria
Sheckler, RIAA); Joint Comments of American Association of Independent Music et al., at 25-26, In re Section 512
Study: Notice and Request for Public Comment (No. 2015-7), available at
https://www.regulations.gov/contentStreamer?documentId=COLC-2015-001389806&attachmentNumber=1&contentType=pdf.

6

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

incentive to file notices. 22 And third, the vendor ecosystem has matured; third parties that handle
generation and submission of notices have become more sophisticated and automated, driving up
submission numbers.
Additionally, a significant portion of the recent increases in DMCA submission volumes for Google
Search stem from notices that appear to be duplicative, unnecessary, or mistaken. As we explained at the
San Francisco Roundtable, a substantial number of takedown requests submitted to Google are for URLs
that have never been in our search index, and therefore could never have appeared in our search results. 23
For example, in January 2017, the most prolific submitter submitted notices that Google honored for
16,457,433 URLs. But on further inspection, 16,450,129 (99.97%) of those URLs were not in our search
index in the first place. Nor is this problem limited to one submitter: in total, 99.95% of all URLs
processed from our Trusted Copyright Removal Program in January 2017 were not in our index.
In order to ensure that these URLs never appear in our search results in the future, we do accept notices
for URLs that are not in our search index. Nevertheless, these “hypothetical” DMCA notices should not
be assumed to correspond to a large number of suspicious URLs in Google’s search results. They instead
reflect the fact that submitters file notices for URLs without regard to whether those URLs are in
Google’s index, and thus could ever have turned up in search results. In fact, of the 35,000,000 URLs we
processed in the latter half of September 2016 that were not in our index, fewer than 2% of those would
have made it into our index in the intervening four months; notices for the other 98% therefore were at
best unnecessary. The large number of takedown requests Google receives is thus not a good proxy for
the number of allegedly infringing URLs that appear in Google’s index.
Nor is the large number of takedown requests to Google a good proxy even for the volume of infringing
material available on the Internet. Many of these submissions appear to be generated by merely
scrambling the words in a search query and appending that to a URL, so that each query makes a different
URL that nonetheless leads to the same page of results.24 Other websites report similar experiences. 25

22

Google, Inside Search (Aug. 10, 2012), available at https://search.googleblog.com/2012/08/an-update-to-oursearch-algorithms.html.
23

Transcript of May 13, 2016 Section 512 Public Roundtable, at 77-78, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-13-2016.pdf (comments of Fred von
Lohmann, Google).
24

See, e.g., DMCA (Copyright) Complaint to Google (Aug. 29, 2016), archived at
https://lumendatabase.org/notices/12937354#.
25

4shared: Copyright Holders Abuse Google’s DMCA Takedown System, TorrentFreak (Nov. 23, 2016), available
at https://torrentfreak.com/4shared-copyright-holders-abuse-googles-dmca-takedown-system-161123/ (“What we
can see is that numerous complaints provide a redundant volume of links that look like some machine-built template
as well as a large amount of non-informative links to various parts of the 4shared website... .”); see also Pirate Site
with No Traffic Attracts 49m Mainly Bogus DMCA Notices, TorrentFreak (Feb. 19, 2017), available at
https://torrentfreak.com/pirate-site-with-no-traffic-attracts-49m-mainly-bogus-dmca-notices-170219/ (“Instead of

7

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

Looking beyond Google’s own experience, most OSPs experience a “relatively infrequent” volume of
notices. 26 For these OSPs, the most comprehensive study on the topic has concluded that the notice-andtakedown process is working well: non-automated “notice and takedown is still the most common
practice … , and, for most, it is still sufficient to manage the takedown requests they receive.”27 Given
this success, the study concluded that requiring “more expensive measures risks undermining the essential
success of notice and takedown in supporting a robust, competitive marketplace for online speech
platforms.” 28
Another sound metric for measuring the overall effectiveness of the DMCA safe harbors is to examine the
diversity of voluntary measures being developed and deployed by OSPs to go beyond the requirements of
Section 512. Congress emphasized that the DMCA was designed to foster collaboration and innovation in
the battle against copyright infringement online, and, as explained here and in our April 1 Comments, it
continues to succeed when measured against that standard. The list of voluntary measures deployed by
various OSPs is long, with new strategies emerging and being refined by many market participants. In
addition to the automated content recognition systems mentioned above, and the many Google-sponsored
efforts described in our April 1 Comments, other OSPs are innovating with “follow the money” efforts,
user education campaigns, and community norm-setting practices. 29 While stakeholders can be expected
to disagree about the details of these voluntary efforts, it cannot be said that the DMCA safe harbors are
failing in the face of this overwhelming evidence that these voluntary measures continue to grow both in
number and diversity.
Question 4: Several public comments and roundtable participants noted practical barriers to
effective use of the notice-and-takedown and counter-notice processes, such as differences in the
scanning the site and sending an accurate takedown notice to Google, APDIF tries to guess the URLs where
MP3Toys stores its content. … APDIF makes up its own URLs, MP3Toys randomly generates a page of music that
has nothing to do with the URL input, APDIF logs it as an infringement of its clients’ rights, and sends a complaint
to Google.”).
26

Jennifer Urban, Joe Karaganis & Brianna Schofield, Notice and Takedown in Everyday Practice, at 2 (U.C.
Berkeley Pub. Law Research, Working Paper, No. 2755628, 2016), available at https://ssrn.com/abstract=2755628.
27

Id. at 113.

28

Id. at 119.

29

See Transcript of May 13, 2016 Section 512 Public Roundtable, at 59-60, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-13-2016.pdf (comments of Charles
Roslof, Wikimedia Foundation) (describing volunteer efforts and community standards relating to policing
Wikimedia sites for potentially infringing content); Comments of Wikimedia Foundation, at 6-7, In re Section 512
Study: Notice and Request for Public Comment (No. 2015-7), available at
https://www.regulations.gov/contentStreamer?documentId=COLC-2015-001388846&attachmentNumber=1&contentType=pdf (same); Copyright School, YouTube, available at
https://www.youtube.com/copyright_school; White House Office of the U.S. Intellectional Prop. Enf’t Coordinator,
Best Practices Guidelines for Ad Networks to Address Piracy and Counterfeiting, available at
http://www.2013ippractices.com/bestpracticesguidelinesforadnetworkstoaddresspiracyandcounterfeiting.html.

8

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

web forms used by ISPs to receive notices or adoption by ISPs of additional requirements not
imposed under the DMCA (e.g., submission of a copyright registration or creation of certain web
accounts). What are the most significant practical barriers to use of the notice-and-takedown and
counter-notice processes, and how can those barriers best be addressed (e.g., incentives for ISPs to
use a standardized notice/counter-notice form, etc.)?
There now exists a robust ecosystem of third parties that have expertise in sending takedown requests.
Similarly, the Copyright Alliance’s recent survey of creators asks whether it should become such a
clearinghouse. As mentioned above, several vendors already offer services navigating YouTube’s Content
ID.
Nevertheless, during the roundtables, we heard smaller rightsholders express concerns about the obstacles
they face in accessing and making the best use of the kinds of sophisticated online enforcement vendors
and tools that larger rightsholders rely on today. To address these barriers, the Office could convene a
multistakeholder discussion about ways to allow smaller entities to take advantage of tools voluntarily
developed by larger-scale entities to address any impediments in notice-and-takedown procedures,
perhaps through collaboration with intermediaries or clearinghouses. It may be that by aggregating
demand for these services, intermediaries like the Copyright Alliance can serve a valuable role,
developing both the strategic expertise to advise smaller rightsholders about the requirements and
capabilities of available enforcement tools, as well as aggregating enough supply-side demand to reach a
scale that will attract the enforcement vendors that have traditionally served larger rightsholders. OSPs
and Internet users, for their part, could benefit as well, insofar as intermediaries who develop expertise
with existing tools are less likely to misunderstand or misuse those tools.
In contrast, mandating standardized practices for OSPs would hinder efficient processing of notices.
Different OSPs have different processes, as befits the enormous range in the volume of notices and the
size of the OSPs’ businesses. Attempts to ignore this diversity, and to curtail flexibility, will only create
problems.
We also must take the opportunity to clear up some misinformation about Google’s own submission
process. One commenter at the New York Roundtable aimed a number of criticisms at Google’s process
in particular. 30 He complained that use of Google’s DMCA forms requires a Google account, which is
true. Requiring a Google account reduces abusive notices from anonymous users and also allows us to
show submitters a dashboard displaying the status of their previous notices, a feature that was requested
by many rightsholders. And, of course, we continue to receive notices through several other channels that
do not require a Google account. The commenter also pointed out that we explain at the appropriate step

30
Transcript of May 2, 2016 Section 512 Public Roundtable, at 71-72, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-02-2016.pdf (comments of Stephen
Carlisle, Nova Se. Univ.).

9

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

in our form that merely being the subject of a photo does not give one a copyright interest in the photo. 31
In our experience, this warning dramatically cut down on the number of misguided notices. This, in turn,
streamlined the removal process for meritorious notices and decreased our turnaround time for removal of
images. Another participant in that roundtable claimed that Google “until recently” made it impossible to
paste a URL into our DMCA web form. 32 That is incorrect. While we welcome feedback about our own
process, which we continue to refine, none of these issues rises to the level of requiring legislative
attention, and even if they did, what works for large OSPs like Google would not be the right solution for
smaller OSPs.
Question 6: Participants also noted disincentives to filing both notices and counter-notices, such as
safety and privacy concerns, intimidating language, or potential legal costs. How do these concerns
affect use of the notice-and-takedown and counter-notice processes, and how can these disincentives
best be addressed?
Google remains concerned that the existing counter-notice remedy has not prevented abuses of the noticeand-takedown process. 33
Additional recent examples of inappropriate takedown notices include:

31

●

A man seeking to remove from Google’s search index the DOJ press release announcing his
guilty plea. 34

●

A social media monitoring service seeking to remove its brochure from Scribd after its activities
were the subject of a critical report by the ACLU.35

●

A lawyer seeking removal of a blog post criticizing the lawyer for plagiarizing content on his
website. 36

Id.

32

Transcript of May 2, 2016 Section 512 Public Roundtable, at 25, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-02-2016.pdf (comments of Eugene
Mopsik, Am.Photographic Artists) (“A friend of mine who engages regularly in notice-and-takedown said that until
recently, in trying to get notice-and-takedown with Google, when they coded their field for putting in the URL, they
wouldn’t allow you to paste.”).
33

See Google’s April 1 Comments, at 11 (identifying examples).

34
DMCA (Copyright) Complaint to Google, Lumen (Nov. 25, 2016), available at
https://lumendatabase.org/notices/13439312#.
35

Mike Masnick, Geofeedia, In Damage Control Mode, Issues Bogus DMCA over Brochure Posted by Reporter,
TechDirt (Oct. 21, 2016), available at https://www.techdirt.com/articles/20161019/00080135834/geofeedia-damagecontrol-mode-issues-bogus-dmca-over-brochure-posted-reporter.shtml.
36

Mike Masnick, Questionable DMCA Takedown Notice Filed over Post Calling Lawyer Out for Copyright
Infringement, TechDirt (Jul. 11, 2016), available at

10

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

●

A lawyer seeking removal of a document revealing his salary from Google’s search index. 37

●

A person posing as a representative of a police department filing DMCA notices demanding that
Google remove from the search index mugshot photographs included in news stories related to
arrests. 38

Other recent examples of inappropriate notices illustrate the challenges of identifying ownership and fair
use.
●

A movie studio issued a takedown notice for a “neural network” experiment related to Blade
Runner, plainly a fair use. 39

●

A movie studio issued a takedown notice for public source software that it indisputably does not
own. 40

Potential Future Evolution of the DMCA Safe Harbor System
Question 12: Several study participants have proposed some version of a notice-and-stay-down
system. Is such a system advisable? Please describe in specific detail how such a system should
operate, and include potential legislative language, if appropriate. If it is not advisable, what
particular problems would such a system impose? Are there ways to mitigate or avoid those
problems? What implications, if any, would such as system have for future online innovation and
content creation?
Google has addressed this question previously in both our April 1 Comments and at the San Francisco
Roundtable. We continue to believe that while “staydown” technologies can play a valuable role in both
preventing infringement and fostering new business models (as Content ID does), they are best developed
in the context of voluntary efforts between rightsholders, OSPs, and user communities. A legal mandate
https://www.techdirt.com/articles/20160711/11004234937/questionable-dmca-takedown-notice-filed-over-postcalling-lawyer-out-copyright-infringement.shtml.
37

Mike Masnick, Sony Pictures Legal Affairs VP Files Bogus DMCA Notice Because His Salary Is Listed on
Wikileaks, TechDirt (Jul. 11, 2016), available at
https://www.techdirt.com/articles/20160711/01191434930/sony-pictures-legal-affairs-vp-files-bogus-dmca-noticebecause-his-salary-is-listed-wikileaks.shtml.
38

Mike Masnick, Burlington Police Insist Someone Is Pretending to Abuse Copyright Law to Censor News Stories
About Arrests, TechDirt (Nov. 21, 2016), available at
https://www.techdirt.com/articles/20161121/00395236095/burlington-police-insist-someone-is-pretending-to-abusecopyright-law-to-censor-news-stories-about-arrests.shtml.
39

Aja Romano, A Guy Trained a Machine to “Watch” Blade Runner. Then Things Got Seriously Sci-Fi, Vox (Jun.
1, 2016), available at http://www.vox.com/2016/6/1/11787262/blade-runner-neural-network-encoding.
40

Mike Masnick, Another Day, Another Anomaly: Paramount Issues DMCA Takedown on Ubuntu Linux Torrent,
TechDirt (Sept. 13, 2016), available at https://www.techdirt.com/articles/20160912/09295735500/another-dayanother-anomaly-paramount-issues-dmca-takedown-ubuntu-linux-torrent.shtm.

11

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

imposing “notice-and-staydown” on OSPs, in contrast, would not be advisable.
We have already described the technical feasibility issues and fair use equities that would make a
“staydown” legal mandate on all OSPs impossible to implement. 41 Comments from a wide range of OSPs
in this Study, moreover, make it clear that the tools to implement a “staydown” obligation are often
nonexistent or too expensive for most OSPs. Nor has there been any evidence submitted to suggest the
existence of any difficulty with the DMCA’s notice-and-takedown approach as applied to the vast
majority of OSPs who rely on the DMCA safe harbors. In fact, research submitted by Professors Urban,
Karaganis, and Schofield suggests that for the vast majority of so-called “DMCA Classic” providers, the
notice-and-takedown process is working well for service providers and rightsholders alike. 42
This is not to say that the DMCA’s notice-and-takedown regime should be viewed as a “ceiling” on OSP
efforts. In fact, the record developed in this Study establishes that many OSPs have undertaken voluntary
measures to go well beyond the requirements of the DMCA safe harbors. Some of those voluntary
measures have included automated “staydown” systems, where technically feasible, appropriate for the
platform, and justified by the scope of the problem. These voluntary efforts are entirely consistent with
the DMCA’s stated goal of fostering collaboration between rightsholders and service providers.
It is difficult to see how a legislative mandate for “staydown” could improve this state of affairs. For the
OSPs that have already moved ahead with these technologies on a voluntary basis, the legislative mandate
would be redundant and constraining, reducing the flexibility to experiment with different measures
where appropriate. For the “DMCA Classic” OSPs, for whom notice-and-takedown is a well-functioning
response to the scale of the infringements on their platforms, a legislative “staydown” mandate would
impose huge new costs for no significant benefit to rightsholders. And for the rogue sites operating
beyond the reach of the DMCA, legislative “staydown” mandates would be a meaningless gesture (even
as other voluntary measures are proving effective).43
Other Developments
Question 14: Several study participants mentioned concerns regarding certain case law
interpretations of the existing provisions of section 512. Additionally, two new judicial decisions
have come out since the first round of public comments was submitted in April 2016. What is the
impact, if any, of these decisions on the effectiveness of section 512? If you believe it would be
appropriate to address or clarify existing provisions of section 512, what would be the best ways to
address such provisions (i.e., through the courts, Congress, the Copyright Office, and/or voluntary
41

See Google’s April 1 Comments, at 10.

42

Jennifer Urban, Joe Karaganis & Brianna Schofield, Notice and Takedown in Everyday Practice, at 114-15 (U.C.
Berkeley Pub. Law Research Paper No. 2755628, 2016), available at https://ssrn.com/abstract=2755628.
43

A Bunch of Weak Anti-Piracy Measures Are Still a Pest to Pirates, TorrentFreak (Jan. 7, 2017),
https://torrentfreak.com/a-bunch-of-weak-anti-piracy-measures-are-still-a-pest-to-pirates-170107/.

12

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

measures)? Please provide specific recommendations, such as legislative language, if appropriate.
The Second Circuit’s decision in Capitol Records, LLC v. Vimeo LLC, 826 F.3d 78 (2d Cir. 2016),
correctly interprets section 512(c) and the protections it provides to both service providers and
rightsholders. First, the Second Circuit was correct to read section 512(c) to encompass pre-1972 sound
recordings. As the Capitol Records court noted, Congress crafted a careful compromise in section 512(c)
between promoting the growth of the Internet economy on the one hand, and preventing the infringement
of copyrighted material on the other. Requiring service providers to police user content for state copyright
law violations, while alleviating the burden of doing so for federal copyright violations, would eviscerate
that compromise and force service providers to “incur heavy costs of monitoring every posting to be sure
it did not contain infringing pre-1972 recordings, or incur[] potentially crushing liabilities under state
copyright laws.” Id. at 90. This would defeat the statute’s purpose and “either dissuade service providers
from making large investments in the expansion of the growth and speed of the Internet (which Congress
sought to encourage) or perhaps cause them to charge so much for the service as to undermine
substantially the public usefulness of the service Congress undertook to promote.” Id. at 92-93.
Second, the Capitol Records Court correctly interpreted the safe harbor’s “red flag” provision to preclude
liability where a service provider’s employee merely viewed some or all of a video that included a
copyrighted work. As the court sensibly acknowledged, such a viewing could be for many different
reasons, including technical assessment, or the detection of inappropriate obscenity or bigotry; it could be
done with the sound on or muted. Id. at 96. Moreover, determining whether a given video, image, song, or
text is infringing requires background knowledge about, among other things, whether the item is protected
by copyright, who owns the relevant copyrights, who posted the item, the poster’s relationship with the
rightsholder(s), whether one of the copyright owners licensed or otherwise approves of the use of the
content, and whether that use is fair or otherwise permitted by law. The Capitol Records Court rightly
understood that a service provider’s employee cannot be expected to know or understand each of those
factors, and therefore, cannot be found to have “red flag” knowledge based simply on a viewing. The
Capitol Records Court properly interpreted section 512(c) in so holding.
Question 15: What approaches have jurisdictions outside the United States taken to address the
question of ISP liability and the problem of copyright infringement on the Internet? To what extent
have these approaches worked well, or created problems for consumers, content creators, ISPs, or
other stakeholders?
The DMCA’s Section 512 safe harbor regime has proven to be the blueprint for a growing international
consensus with respect to how to address the role of OSPs in countering copyright infringement on the
Internet. The notice-and-takedown and “no monitoring” concepts, in particular, were adopted by the
European Union in the E-Commerce Directive shortly after the DMCA became law in the United States.
The core structure of Section 512 (notice-and-takedown and no obligation on OSPs to monitor) has also
been mirrored in numerous multilateral and bilateral trade instruments, championed in particular by the
U.S. Trade Representative. The U.S. has included these provisions in trade agreements with Australia,
13

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

Bahrain, Chile, Colombia, Korea, Morocco, Oman, Panama, Peru, and Singapore. Similar provisions have
been adopted by numerous other countries as well, including Japan, Brazil, and New Zealand. 44
The global nature of the Internet means that the ability of OSPs to do business is fostered by increasing
international legal harmonization, including in the area of copyright. Given the international adoption of
the norms set forth in Section 512, as well as the success of the U.S. Internet sector in the global
economy, we urge the Office to be cautious before recommending any substantial deviation from this
established transnational legal norm. Reliance on the global applicability of provisions in U.S. law, such
as Section 512, has enabled the U.S. to generate a significant annual trade surplus in digitally deliverable
services of over $158 billion. 45 Pulling the rug out from under this global standard would put this trade
surplus at risk and damage U.S. digital leadership. Furthermore, a demonstration that the U.S. is open to
tinkering with core components of Section 512 may inadvertently inspire other countries to do the same—
tinkering with their industrial policy frameworks to increase regulation, liability, and outright blockage of
U.S. OSPs. These approaches would harm not only U.S. Internet providers, but also millions of U.S.
consumers and hundreds of thousands of U.S. small businesses and content creators that leverage OSPs to
reach global customers.
As the Office is aware, several countries are actively seeking to develop frameworks for dealing with
online infringement. For example, the European Union is currently considering a proposal as part of its
“Digital Single Market” proceeding that would impose a filtering mandate on online hosting providers.
The proposal has been widely criticized as ill-advised and unworkable, for many of the same reasons that
have been aired during this Study. 46
Other countries lack any intermediary liability protections at all, while some countries such as Ukraine
and Vietnam are considering or maintaining approaches that are inconsistent with U.S. values around an
open Internet. 47 Given the rise of overbroad liability or monitoring measures that threaten the growth of
44

See generally Stanford Ctr. for Internet & Society, World Intermediary Liability Map, available at
http://cyberlaw.stanford.edu/our-work/projects/world-intermediary-liability-map-wilmap.

45

U.S. Dept. of Commerce, Econs. & Statistics Admin., Digitally Deliverable Services Remain an Important
Component of U.S. Trade (May 28, 2015), available at http://esa.doc.gov/economic-briefings/digitally-deliverableservices-remain-important-component-us-trade.
46

See, e.g., European Copyright Society, Answer to EC Consultation on the Role of Publishers in the Copyright
Value Chain, at 7 (Jun. 15, 2016), available at
https://europeancopyrightsocietydotorg.files.wordpress.com/2016/06/ecs-answer-to-ec-consultation-publishers-rolejune16.pdf (raising concerns that the proposal “will distort competition in the emerging European information
market”); Sophie Stalla-Bourdillon et al., A Brief Exegesis of the Proposed Copyright Directive (Working Paper,
2016), available at https://papers.ssrn.com/sol3/papers2.cfm?abstract_id=2875296; Sophie Stalla-Bourdillon et al.,
Open Letter to the European Commission - On the Importance of Preserving the Consistency and Integrity of the EU
Acquis Relating to Content Monitoring within the Information Society (2016), available at
https://papers.ssrn.com/sol3/papers2.cfm?abstract_id=2850483.
47

See Comments of Internet Association, at 12, 17-18, 2017 Special 301 Review, available at

14

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

the global Internet, and the centrality of Section 512 to the growth of the U.S. economy, this is a critical
time for U.S. leadership to support and promote Section 512 in the U.S. and abroad. 48
Question 16: Please identify any other pertinent issues that the Copyright Office may wish to
consider in conducting this study.
Google would like to take this opportunity to correct the record on several issues that came up in others’
written comments and at the roundtables. We offer these corrections respectfully, in the hope that more
accurate information will lead the U.S. Copyright Office to a better result in its effort.
●

Contrary to a statement made by Fox during the May 12, 2016 Roundtable in San Francisco, 49
Google does employ a hash matching system on Google Drive to prevent further public sharing
of files for which we have received valid takedown notices. 50

●

Contrary to a statement made by the Digital Media Licensing Association during the May 3, 2016
New York Roundtable, 51 Google Image search does have a warning on its results pages that
“Images may be subject to copyright.”

●

Jonathan Taplin suggested at the May 12, 2016 San Francisco Roundtable that YouTube was
exploiting The Band’s anthem “The Weight” without properly compensating legendary musician
Levon Helm. 52 Our records show that the relevant rightsholders have claimed the song, and that
YouTube is paying royalties on it. 53

https://www.regulations.gov/document?D=USTR-2016-0026-0046; Comments of Computer & Communications
Industry Association, at 7, 12, 2017 Special 301 Review, available at
https://www.regulations.gov/document?D=USTR-2016-0026-0034.
48

See Markham Erickson & Sarah Leggin, Exporting Internet Law Through International Trade Agreements:
Recalibrating U.S. Trade Policy in the Digital Age, 24 Catholic U. J. L. Tech. 317, 343-49 (2016).

49

Transcript of May 12, 2016 Section 512 Study Public Roundtable, at 167, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-12-2016.pdf (comments of Elizabeth
Valentina, Fox Entm’t Grp.).
50
Google Drive Uses Hash Matching to Detect Pirated Content, TorrentFreak (Feb. 11, 2017), available at
https://torrentfreak.com/google-drive-uses-hash-matching-detect-pirated-content/.
51

Transcript of May 3, 2016 Section 512 Study Public Roundtable, at 74, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-03-2016.pdf (comments of Nancy
Wolff, Dig. Media Licensing Ass’n).
52

See Transcript of May 12, 2016 Section 512 Study Public Roundtable, at 337-38, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-12-2016.pdf (comments of Jonathan
Taplin, Annenberg Innovation Lab, Univ. of S. Cal.).
53

Google is not in a position to look behind music industry contracts allocating rights to musical works when it
distributes Content ID royalties. For more on The Band and Levon Helm’s personal tragedies, see Allen St. John,
The Hard-Won Lessons of Levon Helm’s Life in the Record Business, Fortune (Apr. 19, 2012), available at

15

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

●

A music industry company alleges in its written comments that it devoted “multiple full-time
equivalent staff members” to removing a track by B.o.B from YouTube. 54 Our records show only
34 notices using our automated tools, comprising 109 videos, between 2010 and 2015, and only
857 manual claims of that track in the last seven years. This represents only 1.5% of the total
number of claims, showing that Content ID is properly identifying the vast majority of this
content.

●

The same company alleges it made a similar effort to remove “A Head Full of Dreams” by
Coldplay. Of the 1,887 videos blocked by vendors using our automated tools, more than half were
already subject to a Content ID claim for the music. This means the vendor’s block caused a
rightsholder that already had chosen to monetize to lose out on future revenue from that claim.

●

The Music Community submission repeats Cary Sherman’s assertion that vinyl sales generated
more revenue “for the industry” than ad-supported streaming in 2015. 55 This comparison
overlooks significant differences between these two forms of content delivery. 56

●

The Content Creators Coalition commented at the May 13, 2016 San Francisco Roundtable that
Content ID’s terms require a creator to license its entire back catalogue to YouTube in order to

http://www.forbes.com/sites/allenstjohn/2012/04/19/the-real-world-business-lessons-of-levon-helms-hardscrabblelife/#47558a926f73 (quoting Helm: “When The Band came out we were surprised by some of the songwriting
credits. In those days we didn’t realize that song publishing—more than touring or selling records—was the secret
source of the real money in the music business. ... Back then we’d get a copy of the album when it came out and
that’s when we’d learn who’d got the credit for which song.”); John Barry, A Statement From Michael Pinsky,
Attorney for Levon Helm, Poughkeepsie J. Sonic Storm Blog (Oct. 28, 2007), available at
http://blogs.poughkeepsiejournal.com/sonicstorm/2007/10/28/a-statement-from-michael-pinsky-attorney-for-levonhelm/ (“The failure of record labels, movie, publishing and production companies to pay Levon and his Band mates
their fair performance and publishing royalties added insult to injury during this time when every dollar counted.
(Record and movie deals negotiated by The Band’s trusted lawyers and accountants, in retrospect, seem designed to
work for the entertainment companies and not for the artists.)”).
54

Comments of Warner Music Group, at 8, In re Section 512 Study: Notice and Request for Public Comment (No.
2015-7), available at https://www.regulations.gov/document?D=COLC-2015-0013-86022.
55

Joint Comments of American Association of Independent Music et al., at 3 & n.6, In re Section 512 Study: Notice
and Request for Public Comment (No. 2015-7), available at
https://www.regulations.gov/contentStreamer?documentId=COLC-2015-001389806&attachmentNumber=1&contentType=pdf.
56

See Timothy Geigner, Despite Massive Streaming Revenue Gains, RIAA Still Lying & Crying, TechDirt (Mar.
24, 2016), available at https://www.techdirt.com/articles/20160323/10074433993/despite-massive-streamingrevenue-gains-riaa-still-lying-crying.shtml (“First, retail value isn’t actual revenue, as it ignores the reality of
discounts and other pricing differences. Second, and more importantly, you’re leaving out the fact that much of the
money from retail value doesn’t go to the labels, but to the retailers, distributors and to cover production costs of the
actual records. Third, with free streaming, you have none of those costs, and the contracts you have with streaming
companies mean a much greater percentage of the revenue actually goes to the labels. Fourth, free streaming does
two useful things that you ignore: (1) moves people away from infringement, where you earn nothing and (2) moves
many people up the chain to paid streaming.”).

16

 Google Inc.
25 Massachusetts Avenue, NW
Washington, DC 20001

Main 202.346.1100
Fax 202.346.1101
www.google.com

make use of Content ID. 57 As Fred von Lohmann testified at the time, 58 and as we reiterate here,
that is untrue.
*

*

*

Google appreciates the opportunity to share its perspective and experience, and we look forward to
continued engagement with the Office on these topics.

57

Transcript of May 13, 2016 Section 512 Study Public Roundtable, at 41-42, available at
https://www.copyright.gov/policy/section512/public-roundtable/transcript_05-13-2016.pdf (comments of Tom
Murphy, Content Creators Coal.).
58

Id. at 78-79 (comments of Fred von Lohmann, Google).

17