Recap of the Copyright Office’s Section 512 Study Roundtable

On Monday, I participated in a Copyright Office roundtable regarding their long-delayed report on Section 512. The roundtable was intended to update the study’s record from 2017, when progress stalled on the report. Thus, the topic nominally was to discuss how the law, especially the caselaw, had changed in the intervening years. In fact, 512 caselaw has noticeably slowed down since 2017, so there are only a handful of cases to discuss, and they reached messy results. Some cases since Jan. 1, 2017: Cox, Grande, Motherless, Mavrix v. LJ (we never blogged it), Zazzle. The 512 haters were aglow about the EU Copyright Directive, and predictably they favored extending it to the US; and several were unrestrained in saying that they disliked Section 512’s structural burden-sharing approach.

I’m including my notes from the day below. As usual, these are my impressions, not verbatim transcriptions of the remarks. In many cases, I didn’t use the exact words of the participants. Please don’t rely on my summaries as authoritative characterizations of what was said. A recording and transcript will be available from the Copyright Office, so I encourage you to consult those. I’ve added limited editorial comments in brackets. Rebecca Tushnet did a more faithful transcription and also added more editorial comments. See her recaps:

Structurally, the roundtable gave each panel participant up to 45 seconds to make introductory remarks, followed by Qs from the Copyright Office representatives. 45 seconds was a ridiculously small time, but most speakers tried not to abuse it.

* * *

Karyn Temple: In past roundtables, it was a tale of two cities on whether the DMCA is working well or not. She hopes to see if that has changed. [spoiler: it didn’t]

Copyright office representatives for the first three panels:

Regan Smith, General Counsel

Brad Greenberg, Counsel for Policy & International Affairs

Kevin Amer, Deputy GC

Kimberley Isbell, Senior Counsel for Policy and International Affairs

Maria Strong, Deputy Director of Policy and International Affairs


Eric Carey, NMPA. BMG v Cox is good, but music industry is still unhappy. Cox involved extreme facts and high enforcement costs. Enforcement system is rigged. Copyright owners shouldn’t bear enforcement burden.

Ken Hatfield, Artist Union. 512 is an unfair loophole that benefits big tech companies.

Mike Lemon, IA. DMCA has right incentives to increase collaboration. The industry relationships keep evolving.

Mickey Osterreicher, NPPA. Rampant misappropriations are taking place online. 512 notices are dangerous and hard. EU obligations could be helpful.

Jennifer Pariser, MPAA. Piracy devastates copyright industry. Cox/Grande are welcome but obvious outcomes. Mavrix and Zazzle are limited to their facts. Notice-and-takedown cases have kept going without acknowledging red flags. It’s all bad news.

Meredith Rose, Public Knowledge. We can’t sever the safe harbors from the entire DMCA scheme. Packingham recognized the importance of getting online, so we shouldn’t cut off broadband access.

Aws Shemmeri, ImageRight. LiveJournal decision is good.

Rasty Turek, Pex. Platforms reject technological solutions.

Rebecca Tushnet, Harvard Law & OTW. Big organizations receive few legitimate takedown notices. KindleWorld takedowns are based on competitive concerns. Don’t write rules to hardwire YouTube practices.

Brian Willen, WSGR. DMCA works. It encourages cooperation. The main burden is on copyright owners. Courts are getting it right. Motherless is a model of DMCA interpretation. Legitimate services are protected by safe harbor, and piratical services are driven out.

Robert Winterton, NetChoice. There is a cottage industry of enforcement services, but platforms have grown to let people express themselves. 512 strikes the right balance. In contrast, Article 13 undermines creativity. We should export 512 around the world.

Pariser: Cox and Grande are right on the repeat infringer issue but those results are obvious. Motherless is mixed bag. The MPAA disagrees that any kind of repeat infringer policy works. Saying that there isn’t a need to track takedown notices doesn’t make sense.

Tushnet: Motherless is a one-person operation. Don’t treat it like YouTube.

Willen: we should consider the nature of the site in determining its obligations.

Smith: is there a minimum standard?

Willen: courts have focused on strikes. In Cox, the consequences of termination were different than termination by web host. Also, we should think about ways to educate users who don’t know the rules.

Smith: should the burden be on copyright owners?

Carey: no

Smith: have cases changed that at all?

Carey: if Cox’s practice didn’t violate, what would?

Greenberg: has pendulum swung?

Carey: we don’t think the balance is right yet.

Hatfield: unreasonable to apply standards from small services to giants. For music, they should filter by ISRC code. Sending takedown notice costs $1,500+; litigation is even more. Musicians make most money in first 18 months of work, but lawsuits don’t resolve during that time period. 512 helps drag cases out. If you profit from my work, share the money with me.

Rose: 512 applies to broadband and web hosts, but the stakes are different.

Amer: In a prior roundtable, IAPs said they rejected takedown notices. Cox changes that. Have practices changed?

Carey: practices have changed, and ongoing litigation on this front.

Isbell: are IAP terminations due to 512 repeat infringer policy “state action”?

Rose: no, but it’s equivalent to state action.

Greenberg: what about voluntary measures?

Rose: there are gradations.

Tushnet: can’t say it’s just private action

Pariser: termination should only be in appropriate circumstances, so this can encompass the nature of strikes and the different types of services. But services need actual policies, not made-up policies. Repeat infringer obligation doesn’t implicate the First Amendment. Termination from one IAP may not prevent any access to Internet except in rare areas.

Smith: what about Mavrix?

Shemmeri: they profited, so they should have paid.

Willen: when do sites curate? LJ ruling implied pre-review ended 512. Motherless made it clear that pre-review is OK. Also, we know from 230 that Congress wanted services to remove and filter, especially remove sexually explicit content off the site. We can’t negate 512(c) if services do that.

Amer: is curation permissible?

Willen: YouTube’s suggested videos are a form of curation that don’t eliminate 512. Every service does curation. It doesn’t make sense to penalize curation. The consequence would be junky sites.

Amer: It’s an administrable rule that direct infringement applies when sites “choose” content.

Willen: In LJ, submissions were pre-reviewed, and the service decided which to publish. If doing that level of ex ante selection of publications puts 512 at risk, that puts pressure on 512. But YouTube/UGC sites let people post what they want, then make those submissions more organized and available. Troubling to risk 512 when services police terrorism.

Isbell: IP is carved out from 230. Does that indicate IP should be different?

Willen: 230 says online services have right to, and are encouraged to, police objectionable content.

Smith: what about encouraging content?

Willen: can’t think of a case where a service made decisions about good/bad content and that negates 512. That policy would be inconsistent with Congress’ goals for 230 and not good public policy. What kind of Internet do we want?

Pariser: she objected to the notion that moderator curating content implies no safe harbor is bad for content. No one is curating for copyright. Services are picking content they like/don’t like for their own reasons, not based on infringement (until takedown notice is sent). If an online service demonstrates that they pick-and-choose files, they should have obligation to do that for infringement as well. This doesn’t matter if it’s done pre- or post-upload—whenever they make choices, they need to accept responsibility for infringing content. They should adopt filters, which aren’t expensive.

Amer: any distinction between Motherless (high level monitoring) and more curated service?

Pariser: there is a continuum of curation.

Amer: did Motherless get it right that screening bad content doesn’t disqualify for 512? In contrast to LJ, which made choices of what to publish.

Pariser: Motherless makes sense in light of current jurisprudence. However, courts are misinterpreting 512. It should always be the case that if a site can control the content on their site, they should be filtering for infringement.

Greenberg: if a site filters for anything, does that mean they are always liable?

Pariser: editing content demonstrates the service’s ability to filter.

Lemon: content moderation is a difficult subject. Most content moderation is driven by user flags, which is similar to DMCA (copyright owners must flag). It’s problematic if efforts to moderate content require perfect filtering.

Smith: what about the Zazzle ruling? Is that case just about physical product?

Lemon: different issue when service markets physical product

Smith: is that different from generating eyeballs for ads?

Lemon: services often rely on automated processes. Even human reviewers make mistakes. Consider the scale. Reddit in 2016-18 saw 700% increase in takedown notices received. It’s unrealistic to expect services to make investments that match YouTube Content ID.

Greenberg: we’ve heard that 512 isn’t one-size-fits-all. Would it make sense to impose the obligations once services reach a certain size?

Lemon: Internet companies ramp up really quickly.

Winterton: filters aren’t inexpensive.

Hatfield: how can we monetize music? How hard is it to register ISRC codes?

Smith: what are standard technical measures?

Carey: we can’t send representative list, and there is no more red flags. Must send URL-by-URL list.

Osterreicher: their group encourages photographers to watermark. Services should be able to recognize and screen for watermarks.

Smith: are standards developing for CMI?

Osterreicher: technology is improving to merge information about the image owner with the image, which will confer knowledge to everyone else.

Greenberg: how would a service know who uploaded the image and with what rights?

Osterreicher: That’s a problem. Start by recognizing watermark, then figure it out.

Tushnet: ability to hash child porn isn’t the same in other content types. YouTube is filtering as well as anyone, yet everyone still complains. OTW terminate users for a variety of reasons, but it only gets 10/yr DMCA takedown notices. Doesn’t make sense to impose liability on services for having a TOS.

Amer: does a watermark trigger red flag knowledge?

Tushnet: OTW doesn’t filter. We don’t want to go to war against our users.

Amer: What constitutes red flags knowledge?

Tushnet: people make mistakes, and people are busybodies who pretend to be the copyright owner.

Amer: what’s difference between actual notice and red flags knowledge? Does red flags knowledge have any meaning?

Tushnet: impossible to say in the abstract. Maybe

Lemon: can’t solve the red flags problem. YouTube sees 300 hours of video every minute. Instagram gets 100M photos/day.

Smith: if you have a filtering database, what are best practices?

Lemon: lots of collaboration between platforms and rightsholders, especially as platforms become rightsholders themselves. Ex: Facebook getting license deals from publishers. The interests are aligning in spectacular ways. There are monthly phone calls to figure out best practices. IA members want to have best experience for users, which is usually professional content. Concerned that going above the statute shouldn’t establish new legal baseline.

Osterreicher: wants watermarks to flag work for additional investigation. Maybe revenue share programs can include photographers.

Turek: Content ID isn’t state of the art.

Willen: red flag is difference between objective (actual knowledge) and subjective knowledge.

Smith: what is red flags knowledge?

Willen: every court has come to conclusion that it’s narrow for good reason. Figuring out copyright infringement is hard and requires context. Watermarking photo doesn’t tell you much. Every photo on Internet is owned by someone. Now what?

Amer: it’s hard to think of an example of red flag knowledge. File title isn’t enough.

Willen: in YouTube, file titles that signaled infringement were actually posted by copyright owner as marketing. Maybe red flag would apply when full-length video uploaded pre-release. Otherwise, all files are likely copyrighted, and there is a huge volume.

Isbell: do ISPs use their own service? Ex: it’s easy to find YouTube videos with just music and lyrics. Ex: Pinterest bookmarks by taking images.

Willen: on YouTube, much of that is licensed. Re Pinterest thumbnails, fair use plays a role.

Pariser: content owners feel like the goalposts keep moving. They can’t send catalog of works to create red flags knowledge.

Shemmeri: copyright notices are stripped from digital files, so that undermines red flags.

Tushnet: don’t presume what services are meant to do. There are 6-7 different YouTubes. OTW gets dynamically generated URLs as takedown notices, which covers too much.

Carey: would love to be able to send a notice that confers red flags knowledge

Hatfield: balance is askew. Onus should not always be on copyright owner.

Osterreicher: tale of two takedowns. Last time, we’re talking past each other. Now, maybe there’s greater recognition of the plight of individual content creators.


Jonathan Band, Library Copyright Alliance. Doesn’t like LJ ruling because 512 depends on moderation. Compliance with EU copyright directive makes that dangerous. The Motherless opinion undid some of the dangerous aspects.

Sofia Castillo, AAP: Companies that rely on infringement as business model shouldn’t get 512. Amendment is required.

Stephen Carlisle, Nova Southeastern. 512 isn’t workable from time and expense standpoint for individual copyright owners. In Google search, he easily found video with static album cover plus lyrics.

Caleb Donaldson, Google: YouTube has paid $6B to copyright owners. Google has best-in-class removal tools.

Ken Doroshow, RIAA: Cases are reading red flags knowledge out of the law. He’s pleased to see repeat infringer rulings, but they were extreme cases that don’t teach much.

Doug Hudson, Etsy: he sees increasing fraud, including phishing, false takedowns, and gaming. Also fraud in counternotices. We need a better 512(f). We also need to simplify DMCA for microbusinesses. Pre-filters aren’t one-size-fits-all, especially when applied to material other than digital files.

Keith Kupferschmid, Copyright Alliance: mass copyright infringements are regular occurrence. ISPs are shielded from liability and encouraged to infringe. Courts added a new requirement of copyright registration. Loss of WHOIS. We’re worse off than 2 years ago.

Art Levy, AIMP. Courts write requirements out of DMCA. Service providers have less incentive to work with copyright owners. DMCA offers no remedy to small copyright owners.

Peter Midgley, BYU Copyright Office: universities are unique in DMCA ecosystem. High burden of processing takedown notices.

Sasha Moss, R Street: Consumers are using legal options. We need to migrate towards those.

Mary Rasenburger, Author Guild: many members do self-publishing. Ebook piracy has blossomed. Notice-and-takedown is an absurd way to deal with piracy. Services aren’t cooperating with rightsowners. Look to EU Directive as model.

Amer: re repeat infringer policies, have recent decisions clarified the law? Are there conflicts among the decisions?

Midgely: recent caselaw hasn’t been as helpful. Only thing that is clear is that not enforcing policy is a problem. How can IAPs implement airtight repeat infringer policy? Many notices purport to be based on 512(c), but really target 512(a) activity. His office often refers takedown notices to honor code office.

Amer: is it possible to provide more clarity to universities?

Midgely: Providing Internet access is not the university’s primary function. It must do a cost-benefit of providing Internet access. If it’s difficult for IAPs in the business to implement repeat infringement policy, it’s even more difficult for non-IAPs. 512(e) is useless but maybe could be salvaged. Also, HEOA requires the university to provide notice to the university community about repeat infringer policy.

Band: libraries are another category of IAP, also not susceptible to standards from Cox or Grande. Many library users have no other Internet access. Some laws require Internet access, and library may be only resource.

Smith: should libraries educate users about repeat infringement?

Band: libraries, especially in higher education, take that seriously.

Isbell: do libraries think they fit in 512(a)?

Band: yes

Greenberg: at last roundtable, IAPs referred to “adjudicated” repeat infringers. Still true?

Band: there are different IAPs with different views. Repeat infringer doesn’t mean alleged infringer, but courts are going in different direction.

Amer: what statutory changes should be made to repeat infringer standard?

Band: no need to change the statute. Everything depends on “appropriate circumstances” language.

Castillo: it wasn’t hard for Cox/Grande to implement a policy. Cox didn’t try. Grande didn’t have a policy at all. Copyright Office should say (1) services should consistently and meaningfully implement a policy, (2) services should keep a log of its actions, and (3) services should prevent re-registration by terminated infringers. 512(i)(1) requires notice to users of repeat infringer policy, and Motherless’ policy didn’t say it.

Midgely: what should service do when it gets conflicting info, such as a user who denies infringement. For the university, it’s unclear how to distinguish “system” from “network” termination.

Donaldson: courts are looking at nature and size of platforms, and that’s good. Repeat infringer policies should have variation because of the amount of resources each platform has.

Doroshow: rightsholders had to send millions of notices before it produced caselaw on repeat infringer policy. Also, terminated users keep coming back because there’s no “know your customer” rule.

Amer: how do rightsowners provide notice to 512(a) services?

Doroshow: RIAA does this.

Rasenberger: Copyright Office should provide best practices guidance for topics like repeat infringer or red flags.

Moss: PTO has hotline for inventors. Copyright Office could offer something similar for rightsowners.

Smith: The Copyright Office’s public information office fields many inquiries.

Kupferschmid: “one size doesn’t fit all” should apply to notices by copyright owners as well. Smaller owners can’t afford to do what RIAA does.

Amer: Re “stored at direction of user” and “no duty to monitor” provisions. What is role of human monitoring?

Band: law is muddled. Content moderation goes way beyond copyright.

Amer: if Mavrix wasn’t “stored at direction of the service,” what would be?

Band: we shouldn’t put platforms in the impossible decision of thinking that undertaking content review will end the safe harbor.

Castillo: screening material shouldn’t end safe harbor. If service is screening for substance and not infringement, it should lose safe harbor. Services have incentives not to look. She disagrees with 512(m) interpretation. It was intended to protect privacy laws like ECPA. It’s been incorrectly interpreted more broadly.

Greenberg: what is the distinction between screening for legal and illegal content?

Carlisle: content moderation is good, and we should encourage it. Sites should require uploaders to represent and warrant that they have the necessary rights. We could ask uploaders if they own content, have proper license, or in public domain.

Donaldson: Google TOS already has this provision. Content ID resolves 98% of disputes on YouTube. If you can recognize the song, so can Content ID. Content Match allows smaller owners to find potential infringement. 400k are using the tool.

Smith: how do we know that content owners accepted Content ID?

Donaldson: music industry is plagued by incomplete data.

Levy: Many independent songwriters don’t have a deal with Google.

Amer: is Content ID more available to smaller rightsowners?

Donaldson: some aggregators represent smaller creators.

Amer: why isn’t Content ID available to everyone?

Donaldson: possibility of misidentification of authorized Content ID users.

Doroshow: if a service is benefiting from infringing content and has tools to screen them out, it should lose safe harbor. There are alternatives to Content ID that are cheaper and more widely available.

Greenberg: what technologies have been added to the ecosystem? Have any become standard technical measures?

Donaldson: Content ID isn’t static. $40M investment in last 3 years.

Hudson: filtering technology can’t pick up all of the long tail. Adding filtering requirement doesn’t solve the problem.

Amer: what about uploading the full work? Why can’t filtering technology pick that up?

Hudson: what if the work is a quilt? Too much focus on works that start out as digital.

Greenberg: what about T-shirt with full movie poster depicted?

Hudson: vintage T-shirt/poster can’t be determined

Smith: has Etsy changed its policy after Zazzle?

Hudson: Etsy is a pure marketplace. No print-on-demand, no drop-shopping. So Zazzle is different.

Kupferschmid: filtering can account for wide range of circumstances. If something gets flagged, the service could ask the user for clarification.

Band: SSRN pre-reviews content to determine which category to place it in

Amer: that sounds like volitional conduct

Band: the uploading is at user’s direction

Amer: but if service has choice to post, that’s volitional

Rasenberger: TOS isn’t enough to verify. Every site should require uploads to declare the legal basis for uploading. Also, what about clear bad actors who hide behind 512? She gave example of Authors don’t have any choice but to litigate, but they can’t afford to do it.

Isbell: Congress never intended to cover bad actors.

Rasenberger: YouTube and Shelter Capital rulings protect bad actors.

Isbell: is the answer to pull back 512 for everyone, or just find ways to exclude bad actors?

Rasenberger: there should be regulations on red flags. Knowledge should not require specific item at specific location.

Moss: uploaders don’t know what fair use is. Upload filters don’t work. Moderator’s dilemma.

Levy: Lenz is major problem to 512(f). Copyright owner must consider fair use. Doing 4 point analysis for each potentially infringing work is too much.

Greenberg: are automated notices are still OK after Lenz?

Levy: final ruling didn’t clarify.

Hudson: platforms can’t make fair use determination.

Carlisle: Fair use is complex. Independent musician can’t figure it out before sending notice.

Amer: statute assumes copyright owners comparatively have the most knowledge. Isn’t that the basic bargain of the statute?

Carlisle: Putting burden on copyright owners is the wrong bargain.

Smith: 512(f) doesn’t apply to honest mistakes.

Carlisle: Lenz case was hard

Smith: the “knowing” requirement helps 512(f) defendants

Carlisle: musicians can be aggressive about asserting their rights, and they think any similarity is infringing

Castillo: the counternotice system protects legal content, and it works. [both Rebecca and I chuckled at that]

Band: courts hold bad guys liable. Sometimes rightsowners make dubious calls about who is a bad guy.

Rasenberger: Amazon uses fingerprinting and they are good at keeping infringers off the site


Richard Burgess, AIMP; not much has changed from 2017. Still want staydown, send lots of notices, can’t block repeat infringers

Alex Feerst, Medium: 10k posts/day. The Moderator’s Dilemma and the problem with red flags.

Devlin Hartline, George Mason: 512 wasn’t intended as notice-and-takedown. Service providers were expected to prevent infringement without input from copyright owners. Red flag knowledge standard was meant for this purpose. Google can index Pirate Bay shows how far things got out of hand. Red flag knowledge is general knowledge of infringement. [smh]

Cathy Gellis, Copia Institute: don’t let Sky is Falling rhetoric dominate. Sky is Rising. What effect of unadjudicated claims of infringement?

Eric Goldman: [see my blog post on Section 512(f)]

Joe Gratz: not many 512 cases, which signals the rough consensus among industry players. The cases show the flexibility of the law.

Jared Polin, FroKnowsPhoto: He makes a living at YouTube and gives away all of his content, and he supports 2 employees.

Tamber Ray, Rural Broadband Association: Often members are the only broadband providers in their community. They struggle with takedowns, and they are terminating users. They would like guidance on repeat infringer policy.

Bob Schwartz, Consumer Technology Association: curation can’t be a disqualifier for safe harbor

Christian Troncoso, BSA: DMCA works largely as intended. Frustration on both sides shows the compromises works.

Kate Tummarello, Engine: startups like DMCA because it reduces risk of ruinous litigation

Ping Wang, freelancer: favors EU copyright directive

Nancy Wolff, Digital Media Licensing Association: Internet would be empty without visual content, but no incentives for collaboration between service providers and content providers. Fingerprinting technology isn’t being used.

Isbell: At last hearing, Gratz said 512(f) needed more teeth, such as statutory damages. Still feels that way?

Gratz: 512(f) doesn’t deter for many kinds of abuses (notices and counternotices). Not many lawsuits because not enough at stake to support litigation. He expects more lawsuits because marketplaces encourage competitive gaming

Isbell: Etsy mentioned more fraud in notices/counternotices

Greenberg: is legal standard the same for notices and counternotices?

Goldman: yes

Smith: if we loosen legal standard on 512(f), won’t that hurt counternotices?

Goldman: counternotices are relatively small part of ecosystem, so we should focus on abusive notices as the principal problem

Polin: YouTube has good counternotice procedure

Wolff: counternotice makes it harder for copyright owners, who are then forced to court. When counternotices are wrong, it hurts copyright owners

Wang: search engine runs ads on my content

Gratz: if people don’t send counternotices, it’s not because of fears about 512(f) liability

Gellis: 512(g) undermines anonymous users. Fewer deterrences for sending 512(a) notices. People are sending notices to create strikes against targeted people

Greenberg: is there evidence of competitive gaming via strikes?

Hartline: bogus takedowns are a problem, but it’s a forest-for-trees problem. Google has been asked to remove 4B URLs, and most notices requesting those removals were legit. Wrong takedowns are a small problem in comparison. Lenz is wrong because it expects senders to have to actively formed good faith belief before sending.

Feerst: Medium engages with senders of non-compliant DMCA notices. A substantial number of senders don’t appreciate the gravity of the legal consequences of sending the notice. Many notices are sloppy, and the senders won’t engage with Medium to fix it.

Smith: the law gives you enough time to go back-and-forth with senders?

Feerst: Medium takes that risk because “expeditiously” isn’t clear. Medium does it because they care.

Burgess: worried about Gellis’ fact claims. Most labels have given up on notice-and-takedown. Content is the golden goose, not the Internet. Rightsowners own the stuff, we’re not responsible for other people’s misuse. Music industry is now 1/3 of what it was 20 years ago.

Polin: I don’t want to see gatekeepers shutting the gates again. I can make money by giving content away for free.

Isbell: what is optimal way to have notice-and-takedown? Sloppy notices and high volume; vs. content creators frustrated they have to use a form rather than sending email

Ray: members are getting notices at email addresses that aren’t the official one; also getting duplicate notices that need deduping.

Wolff: Lenz hasn’t changed landscape. Visual artists always do manual review before sending notices. Many have given up.

Greenberg: Google gets less notices now than at its peak in 2016. Why?

Gratz: vast bulk of content taken down never appeared in Google in first place

Troncoso: no perfect takedown notice system. Diversity of services and content owners. DMCA supports bespoke systems to develop.

Schwartz: burden has to be on rightsholders.

Goldman: much activity is outside of 512(c)/512(f) now: fast lane for rightsholders giving them trusted removal authority that doesn’t constitute a 512(c)(3) notice; and automated filters that kill content before it goes live, which also doesn’t implicate 512(c) or 512(f).

Wang: wants to know how much Google earns from my works.

Hartline: Lenz is wrong. Do rightsholders need to consider all considerations beyond fair use?

Amer: why isn’t fair use part of the legal evaluation?

Hartline: not knowing about fair use isn’t proof of subjective bad faith

Burgess: no artists believe DMCA is working.

Isbell: what about 512(h) and 512(j)? [note: my notes on 512(h) and 512(j) got a little garbled because Isbell initially referenced a Strike 3 ruling about 512(i), when it actually involved 512(h) [this case], and I multi-tasked to figure out what she was referring to]

Goldman: 512(h) is an anachronism. We don’t see other non-judicially supervised mandatory disclosure of identity in other legal doctrines. It’s supported copyright trolling, where the plaintiffs use it for extrajudicial remedies. It’s also not consistent with modern privacy laws. WHOIS and 512(h) are 1990s solutions, not 2010s solutions.

Isbell: 512(j)?

Goldman: few cases discuss 512(j), but it’s an essential part of the DMCA scheme. 512 doesn’t eliminate liability, it only protects against financial damages and channels injunctions to 512(j)

Gratz: the 512(j) injunction leads to the same outcome as collaboration would produce

Isbell: more comments on repeat infringers?

Tummarello: small companies need certainty/clarity on repeat infringer policy

Schwartz: in Cox, the evidence about DMCA takedown was counted against Cox for contributory infringement

Gellis: determining what constitutes infringement is hard, so counting repeat infringement is imprecise

Smith: not sympathetic to the free expression of infringers in Cox

Gratz: continuum between IAPs and content services. Packingham shows the importance of Internet access.

Greenberg: since Cox, do IAPs now see themselves as subject to same knowledge standards?

Ray: can IAPs ignore emails sent to non-official email address? Grande opinion puts big obligation on IAPs and didn’t consider the balance between valid and invalid notices. Also, what about public wi-fi? FCC and state regulators consider Internet access an “essential service” with rules about termination of service

Feerst: we can’t afford to go to summary judgment on repeat infringer policy.


Joined by Emily Lanza, Office of Policy and International Affairs. Smith and Greenberg stepped off the stage.

Carlo Lavizzari, Internationl STM Association: implementation is key.

Stan Adams, CDT: 512 > Articles 11 & 13

Eric Cady, IFTA. Members don’t have way to stop global infringement. Most are small and medium sized film producers. Encouraged by EU Copyright Directive as way to rebalance notice-and-takedown.

Danielle Coffey, News Media Alliance: wants quality journalism.

Alec French, advocate: giving a safe harbor was an unfortunate pricetag of the rest of the DMCA. Limit DMCA to Internet startups; shouldn’t privilege Internet giants.

Ashley Freedman, ITI: DMCA balances innovation

Josh Lamel, Re:create: Copyright Directive will hurt US startups. Creative industry has increased profits; and piracy is down. Creators are bypassing traditional intermediaries.

Stan McCoy, MPAA Europe: they support no-fault injunctive relief, but the Directive didn’t go enough.

Corynne McSherry, EFF: copyright policy is speech policy and innovation policy. If EU forces upload filters, they will catch lawful non-infringing content. Content ID regularly misidentifies content despite its $100M cost.

Katie Oyama, Google: Directive will harm Europe, especially due to overblocking.

Chris Randle, Facebook: Facebook is taking innovative steps to protect copyright, with the encouragement of DMCA.

Steve Rosenthal, McGraw Hill: WHOIS database is integral to pursuing infringers.

Matt Schruers, CCIA: growth in the creative system.

Lui Simpson, AAP: web blocking is in 40 countries and it hasn’t broken the Internet

Sherwin Siy, Wikimedia: Directive isn’t model law. It’s unclear, novel, uncertain, and internally inconsistent. Wikimedia is large site but small user base. The rules don’t fit unusual sites like this.

Abby Vollmer, GitHub: GitHub got carveout to Article 13, but many other services didn’t.

Rachel Wolbers, Engine: startup exception isn’t workable.

Strong: will EU Copyright Directive affect American businesses doing business in Europe and in US?

Lazzari: does Article 13 codify ECJ caselaw about structurally infringing websites? Two cases are pending (YouTube & Elsevier). Also, he hopes that sites allowing anonymous submissions are still held to higher standard.

Simmons: recital says that it codifies existing law. RapidShare ruling will still be good law

Vollmer: filtering will affect GitHub despite its exception. Open source code often compiles code from many sources, and filtering could detect and break any of them

Siy: it may be impossible to upload works to Wikimedia Commons

Cady: positives: requires license authorizations, and requires staydowns.

Strong: how will authorizations be available?

Cady: members license exclusively, so they may not be willing to license their works to platforms

Lazzari: creates incentive for rightsholders and platforms to cooperate, so works should be available.

McCoy: Berne Convention may prevent formalities before Internet companies are liable. Many members don’t want to license movies to platforms, so this will require enforcement (but maybe filtering will reduce the problem).

Oyama: Directive imposes direct liability. What will constitute sufficient notice? More than 50% of money they pay is based on UGC, and that revenue is at risk.

Schruers: filtering obligations are unmanageable. Only filters are available for AV works, but filtering mandates apply to all works

Simpson: EU will adopt proportionality and reasonableness. Law will “nudge” the creation of filters for non-AV works. [if the EU Copyright Directive is a nudge, I shudder to think what she thinks a punch in the face looks like]

McSherry: copyrights have limitations. Licensing regime overrides those limitations. Filtering is required, but nothing is clear beyond that. Investors are chilled, and the startup exception doesn’t help with that.

Wolbers: Startups look at scalability. Different legal regimes require different platforms. This isn’t feasible; plus filtering cost. Startup carveout creates perverse incentives.

Isbell: EU digital single market was supposed to negate different platforms in EU. Don’t we already have splinternets?

Lamel: Not all creators are the same, and they may want different things. No trade association represents the 17 million people in US who distribute content through platforms completely independently. We should be fighting human rights abuses in other countries’ laws

Coffey: Article 11 creates compilation copyright for news publishers. Local US publishers may be able to benefit from the right.

Adams: startups will conform to most stringent rules, which requires licensing/filtering, which negates fair use.

Lazzari: startups shouldn’t build system that will be victim of its own success. If you want to build UGC, build compliance upfront. But bakeries are safe [it will be a shame to lose UGC on the Internet, but I’m glad we’ll still have baguettes]

Siy: consider privacy implications

Schruers: Directive breaks single digital market

Lanza: how do you advise startups which international laws to comply with and which to ignore?

Schruers: many rules are honored in the breach. Wait for governments to block.

Oyama: criticism of the Thai king is different from determining when works are infringing. Every work is copyrighted.

Strong: over 40 countries have adopted IAP site blocking obligations.

McCoy: loves site-blocking, and its proliferation around the world shows its popularity. It functions well, and none of the predicted dire consequences have come to pass. Good complement to notice-and-takedown; it increases the copyright owner’s toolkit. His only criticism is that some member-states haven’t adopted it.

Simpson: McGraw has pursued site blocking remedy in 6 countries. Rightsowners are careful not to abuse the rights because they are budget-constrained.

Oyama: globally, there are abuses of site blocking. Google does demotion for takedown requests; they have demoted 66k sites. Demotions apply globally.

McCoy: praised Google’s demotion policy.

Simpson: Canada’s notice-and-notice policy does nothing

McCoy: Belgian IAPs are willing to voluntarily block in some cases

Lamel: US IAP market is less competitive than in other countries.

Wolbers: many platforms offer their own private dispute resolution mechanisms

Schruers: diversity of services means that most solutions don’t scale across different sectors.

Simpson: DMCA doesn’t do a good job distinguishing IAPs from other online services.

Strong: repeat infringer policies outside the US?

Simpson: not needed because rightsowners go to court anyway

Isbell: wouldn’t you do both?

Simpson: if service pushes back, go to court.

McSherry: cutting off Internet access to a household has severe consequences to the household.

French: if you’re as big and rich as Google, you don’t deserve DMCA protection

Lamel: copyright policy is looped into broader Internet policy

Vollmer: no one fits into startup exception to Copyright Directive

Amer: complying with DMCA is voluntary, but it provides big benefit. Does the voluntariness reduce the concerns about IAP termination?

Siy: given alternative of strict liability and statutory damages, DMCA safe harbor compliance isn’t really voluntary

Lamel; lack of IAP competition. Also, IAPs and content producers are merging. Both AT&T and Comcast are members of MPAA

McCoy: ECJ has done the work balancing the interest of terminating Internet access


Janice at Rutgers, but speaking as member of public: 512 litigation can’t fix it. There is a permanent conflict between services and rightsowners, and protection increases services’ wealth. Give creative people their rights back. Big companies are backing activists to fight EU directive. Big companies take advantage of flaws in 512. A technical war was waged against EU directive; and Copyright Office was hacked when 512 comments were due. Free speech claims are used to justify ripping off the public.

Kupferschmid: when copyright owners take on burden, they have little to show for it without staydown. They are teaching creators not to send fraudulent notices. CASE Act would enable 512(f) small claims. No defendant could come up with example of red flag knowledge. Fourth Estate case created new DMCA filing requirement [but it’s not new—the statute was always clear that registration was required for enforcement]

Rasenberger: one writer says she spends 50% of her time on enforcement. Does the US want to protect copyright? If so, amend 512. Internet platforms have drained money out of content industries. Shame on us if we don’t fix this. The EU has the courage to fix this. We should too. 512(j) relief is narrow and its application is uncertain. Authors band together to send takedown notices to deliberately demote sites in Google.

Pariser: she wants a separate roundtable just about standard technical measures, because filtering is so ubiquitous. The number of notices being sent is dropping, but that’s not due to reduction in piracy (even though in past roundtables it was claimed that notices rising was a good measure of piracy increases) because of notice-sending fatigue + only sending notices to demote site on Google + shifts of piracy away from P2P/torrents and towards search. Tech companies are making vast money and content is relatively shrinking. Internet services spend a tiny fraction of revenue on anti-piracy efforts

Vicki from RIAA: concern about streamripping. Recording revenues are rising but nowhere near the peak

Hatfield: no one can compete with free. He’s worried about monopsony. Musicians thought they were bypassing record labels, but at least they invested in artists. Musicians just want fair and equitable percentage, and without their content the tech giant platforms will be less valuable.

Goldman: UMG v. Shelter Capital provided a specific example of what constitutes red flags knowledge (third party notices can create such knowledge). It is possible to compete with free—Polin showed us that. Google and Facebook are not the Internet, and regulators are screwing up by enacting anti-Facebook/anti-Google laws that destroy the rest of the Internet. Section 230 has extensive caselaw about what constitutes editorial curation, and it provides an instructive model. Content that starts out as third party content remains third party content regardless of what happens to it, but Batzel excludes content from 230 that was never meant for publication.

[I had to catch my train and missed several other open mic comments. Rebecca’s notes are more complete.]