My Comments to the Copyright Office on DMCA Safe Harbor

Normally, on Plagiarism Today, I don’t involve myself in politics, even when it comes to copyright. My primary goal with this site is to provide practical information and advice for dealing with content misuse of all types, whether it’s online piracy, academic plagiarism or anything in between.

However, the U.S. Copyright Office recently held a public comment period ahead of a study on Section 512 of the copyright act, the section that deals with “safe harbor” provisions of the Digital Millennium Copyright Act (DMCA).

As part of my efforts to deal with plagiarism and infringement, I’ve sent thousands of DMCA notices, both for myself and others through my services at CopyByte. I’ve also been the victim of questionable or even false DMCA notices on some of my YouTube channels.

As such, for the first time ever, I wanted to provide my personal insight and recommendations to the Copyright Office and submitted my comments on April 1st. I used my previous article about DMCA safe harbor as a jumping off point but ended up going in to much greater detail about both the problem and some possible solutions.

The text of my comments are below.

Comments from Jonathan Bailey

Dear Ms. Pallante,

My name is Jonathan Bailey. I have worked for the past 10 years as a copyright and plagiarism blogger at Plagiarism Today and for six years as a consultant at CopyByte. Before that, I was an author using the law to defend my creations as they were being plagiarized and infringed on the web.

Between my job and my own content protection efforts, I have filed thousands of Digital Millennium Copyright Act (DMCA) takedown notices, most of which have been with hosts, and I have successfully used the law to defend both my rights and the rights of my clients, my friends and my family.

To that end, I’ve seen the DMCA at both its best and its worst. Through my work with the Cyber Civil Rights Initiative (CCRI), I have used it to successfully remove revenge and non-consensual pornography from the Internet. However, through my various YouTube channels, I’ve been subject to takedown notices filed on videos with no unoriginal content or content that was a clear fair use.

I’ve seen both the relief and the power of what the DMCA can do when used well but also the frustration of what it’s like to be caught in the crosshairs of a false notice.

It’s clear to me that, after nearly 20 years of life, the DMCA still provides many great benefits and is a powerful tool for many who are infringed. However, those benefits have become tinged by issues with the law’s application in 2016 that could not have been foreseen when the law took effect in 1998.

As such, the DMCA has room for improvement, but it’s important to not throw away the benefits that the law continues to provide and, instead, build upon them to ensure fairness for all of those involved in the DMCA process, including hosts, rightsholders and those who are subject to such notices.

The Arms Race

How effective the DMCA is largely depends upon whom you ask. For smaller rightsholders who see a moderate amount of piracy, the law works reasonably well. They can trivially track their content and file takedown notices to protect their work at a minimal cost of time or money.

Similarly, for smaller online service providers, the law also works well. They receive only a modest amount of notices, the burden of compliance is not onerous and the legal security the law provides far outweighs any drawbacks.

The problem with the DMCA begins when you look at larger rightsholders. Whether record labels, movie studios or software developers, large rightsholders are forced to send out millions of DMCA notices just to keep pace with the rates of piracy they are seeing. 

Nowhere is this more evident than with the Google Transparency Report, which shows that rightsholders are currently sending out over 80 million URL removal requests every month and the volume has been growing rapidly with less than half that number of notices being sent as recently as one year ago.

This DMCA arms race between rightsholders and pirates has resulted in increased automation, both in content detection, the submission of notices and in DMCA compliance. This is seen in the top reporting organizations in the Google Transparency report, which is populated exclusively by companies that provide heavily-automated solutions.

With this automation has come increased mistakes. Machines are simply not as good at detecting infringement and fair use issues as humans. 

However, the extent of this error is extremely difficult to determine.

A recent study claims that nearly 30% of all DMCA notices have some kind of material flaw. While the study itself is questionable, because both the representative sample and the funding for the study were provided by Google as well as the fact that many disqualifiers don’t necessarily mean a DMCA notice is false, there clearly is a very real issue with errors in DMCA notices.

I’ve seen this first hand on YouTube when a video when a video I posted about the TV show Beakman’s World was ordered offline by a DMCA notice when it contained no footage from the show. More recently though we’ve seen the rise of the “Where’s the Fair Use” campaign where YouTuber’s large and small have fought against what they see as questionable takedowns.

However, the issues with YouTube often have less to do with the DMCA and more with YouTube’s in house content detection system, Content ID. Unfortunately, Content ID is just as imperfect the tools rightsholders use and creates many of the same problems. However, those problems are outside of the safe harbor framework.

Still, between Content ID error, rightsholder mistakes and rare instances of maliciously false DMCA notices, many YouTubers feel, understandably bullied. 

However, one attorney who has offered pro bono help to those being abused on YouTube said he received over 700 emails asking for help, but that most weren’t being bullied, they were actual infringers who were unaware that they were in violation of the law. This misunderstanding of the law further complicates trying to understand the DMCA error rate.

In the end, any error rate is too high and, even if it isn’t as high as the recent study or WTFU campaign would lead others to believe, steps should be taken to reduce it.

The Challenges in Fixing It

While a common proposal for reducing error rates is to simply eliminate automation and require all filers to have a human verify each notice, that approach is untenable. Though I personally use no automated tools to file DMCA notices, when rightsholders need to send millions per month just to keep pace with piracy, human verification would require an army and that army would likely be unskilled (thus likely leading to more errors) and/or burdensome in terms of cost.

For some rightsholders, that would make defending their copyrights online impossible.

The focus should instead be on reducing the number of DMCA notices and, thus, the reducing the need for heavy reliance on automated tools. 

One solution that has been offered up is a takedown and stay down system, one where hosts, after receiving a DMCA notice, not only have to remove the work but use tools, similar to Content ID, to prevent the reupload of it.

While that would certainly reduce the number of copyright notices sent, it has its own problems. 

First, not every online service provider is a good candidate for a takedown and stay down system. Though hosts like YouTube can implement them fairly easily, it’s because they host primarily a single type of content and have a single upload point. Web hosts like GoDaddy and Hostgator, however, host almost every type of content and they are uploaded directly to individual accounts rather than processed through a centralized service. 

One possible variation is to have only hosts that profit directly from user-uploaded content be subject to such a system. This would require hosts like YouTube, SoundCloud, most file hosts to implement such a system but spare GoDaddy, Hostgator and others that charge customers to upload rather than running ads or charging for access to user content.

However, that variant would also net Facebook, Twitter, Reddit and other social media services. While such a system would not likely be onerous for the giants, any ad-driven site accepts uploads from users could also be involved, including small forums and other online communities. Smaller sites could struggle to comply with this, even if they see few to no copyright notices per year.

Any solution that involves such a takedown and stay down system should be developed carefully, not just to avoid burdening smaller service providers, but because such a system reintroduces an element of automation, one that blocks future uploads.

Still, it doesn’t mean that there isn’t a place for such a system, just that it must be implemented with great care.

Recommended Improvements to the DMCA

The biggest benefit of the current solution is that it is a one-size-fits-all approach that applies equally to all hosts and rightsholders. For all of its flaws, it adds a great deal of predictability in the relationship between service provider and creator and provides legal certainty for both.

The first step should be to close exploits under the law that allow service providers to do less than what they trivially can to reduce copyright infringement.

1. Require Greater Search Engine Action – Search engines, most notably Google, are only required to remove individual links after a DMCA notice. However, links, files and infringing material reappear quickly, often on the same site but at different URLs. Google has repeatedly promised to take action in demoting heavily-infringing sites but those efforts have failed to produce any real benefit. The same as hosts are required to terminate repeat infringers, so should search engines, that would eliminate much of the current Google DMCA blight as approximately 7.5% of all URL removal requests to Google last month were attributed to just the top 10 domains (11.273 million URLs out of 83,954 million total). Such a system should also deal with sites that pop up under different domains with identical content, such as with proxies and sites that simply move to new domains or hosts.
2. Streamline Submission Process – Though Google processes millions of notices per week, the vast majority of those are through a private, invite-only backend that Google provides larger rightsholders. Smaller copyright holders have to go through a much more onerous 46-step process. Google accepts DMCA notices via email but responds to them slowly, often taking weeks. The law should help streamline the process between creator and service provider. The U.S. Copyright Office can aid this by providing a modern DMCA agent database and facilitating submission of DMCA notices directly, possibly even acting as a middle man. It can also require service providers to respond to notices via email and other, less onerous methods than long, complicated forms some hosts require.
3. Define “Expeditiously”: In my experience, service providers take varying lengths of time to respond to DMCA notices, ranging from mere minutes to weeks. Unfortunately, the law under §512(c)(1)(C) simply says hosts should respond “expeditiously” to a notice, but there is no clarification as to what that means. Even if the law can’t provide a firm deadline, it can and should provide more clarity.
4. Improve Transparency: While some service providers, including Google and Twitter, provide transparency reports, most do not. Requiring hosts who get more than a certain number of notices per quarter to post transparency reports would aid in understanding how the DMCA is being used. However, such transparency reports should NOT include unredacted links to the pirated material, which is what Lumen Database (formerly Chilling Effects) currently does. Google’s Transparency Report is a very good example of providing useful data without linking to infringing material. 

Beyond those changes, the Copyright Office should also consider implementing a takedown and stay down system, but on a voluntary basis.

1. Encourage Stay Down Systems: After tightening requirements on service providers, the law can then offer benefits to hosts who can and do implement qualifying takedown and stay down systems. Those benefits can include greater legal burdens on creators seeking to sue for failure to comply with the law, more time to respond to notices and more control over how notices are sent. This would encourage larger hosts who can implement stay down systems to do without without burdening smaller ones who have no ability and little need to do so.
2. Protect Fair Use: With any stay down system, since it is going to involve automation, there needs to be protections for fair use. While such systems can and should block near-identical uploads, likely fair uses should be allowed up and the system should err on the side of of fair use. However, as a counterbalance, the system should notify the rightsholder to the upload and allow them to challenge it later without requiring them to detect it on their own. If the use is clearly infringing, it should be added to the stay down system. We already see this to some degree with Content ID and it could be implemented elsewhere.
3. Transparency and Appeals: Finally, if a stay down system is implemented, it should be coupled with transparency and appeals. One of the main issues with the YouTube Content ID system is that there is no effective way to appeal an erroneous Content ID match. Users impacted by a stay down system should have a clear set of rights akin to the counter-notice rights under Section 512 currently, albeit with a different structure.

All in all, the goal of any takedown and stay down system should not be to completely eliminate any uploads, but to drastically reduce the reuploads and reduce the burden on both the host and the creator when it comes to dealing with them.

Conclusions

Any changes to Section 512 needs to be carefully measured. While the law certainly has its flaws, for many creators and service providers, it works reasonably well. 

However, the areas where it isn’t working as planned are threatening to endanger the entire process by making it far too onerous for rightsholders and service providers alike and creating a volume of notices that makes mistakes inevitable.

Reducing the number of notices while maintaining the efficacy and simplicity of the system should be the first priority. Doing so will reduce the burden on creators and, in the long run, on service providers. It will also reduce the number of mistakes by making greater human involvement in the submission of notices more practical.

While a takedown and stay down system does not eliminate the use of automation, having it on the provider’s side rather than the rightsholder side gives uploaders greater protection and reduce burdens on everyone as service providers don’t have to process countless DMCA notices and providers don’t have to submit them. Proof of this is found in Content ID, which was implemented by YouTube to streamline the copyright process and is widely participated in by large copyright holders.

All in all, reducing the number of notices will make everyone happy. But it should only be done in a way that doesn’t place undue burdens on smaller service providers, protects fair use and ensures that the law remains effective for those who have benefited both from its protections and its ease of enforcement.

Thank you very much for your time and consideration of these comments and for the opportunity to express my opinions on the subject.

Sincerely,

Jonathan Bailey
Plagiarism Today & CopyByte