Monday, May 02, 2016

Copyright Office 512 Roundtable: Service provider response

Session 2: Notice-and-Takedown Process—Service Provider Response and Counter-Notifications
 
Official description: Handling of and response to notices, including timing and notices from high-volume senders; sending and handling of counter-notifications; volume of notices and counter-notifications; costs and burdens on large- and small-scale service providers; role of automation, including filtering technologies; noncompliant notices and misuse of notice process; rejection of notices; impact of takedowns on users and public; protections for fair use; relevant empirical data; and other pertinent issues. 
 
Claggett: ISP experience: Is it working effectively? Automated notices?
 
Michael Weinberg Shapeways: High level concept works reasonably well given the circumstances. Notice quality is highly variable, which doesn’t track to the size of the content owner. People assume 512 notice and takedown model for complaints unrelated to ©.
 
Claggett: Can you handle notice volume?
 
Weinberg: notices are bursty for us: week w/few and then w/lot. That’s problematic from staff allocation standpoint.  But we can handle it.  As we see #s go up, need to build internal tools.
 
Actual volume: just under 1000 notices last year; trend to go significantly beyond that. One problem is that those notices tend to be an IP grab bag.  Due to our our site’s capacities: 3D printing.  Addit’l burden of processing in ID’ing real complaint.  Easy to forget that burden from 512 standpoint.
 
JC: would it be helpful to have a clear form that maybe said “for copyright claims” on the top—would that aid you?
 
Weinberg: when we tried to update registration w/CO, took 4 months to find out what to do if we didn’t have a fax.  Structure of the form, connecting to back end would have to involve over time; not clear there are mechanisms to evolve a standardized form even if it was correct at the time it was deployed.  On the flip side, if data about notice process became more public, that could be worth the process.
 
Rebecca Tushnet Organization for Transformative Works
The over 600,000 registered creators on our site who’ve posted over 2 million works, and our small all volunteer coding and legal teams, appreciate the opportunity to participate.  Most ISPs are like us, Wikimedia, Internet Archive: we receive relatively few notices, they’re generally illegitimate attempts to assert rights over titles or fair uses, we hand review each one, no capacity to build filters that are unnecessary anyway.
 
JC: Might it make sense for big sites?
 
RT: not here to say that. What I learned from comparing UMG, Sony, and Warner’s comments with those of other copyright owners advocating for increased duties for ISPs: I learned that the best filtering system is the one you’re not using.  UMG says Content ID is 60% effective, Sony says it missed millions of infringements.  The other content companies, like the book publishers, are sure that Content ID must be fantastic and therefore should be applied to the entire internet. But the biggest users of Content ID think it’s full of holes. There’s always something more to demand: like Angelica Schuyler, they will never be satisfied. 
 
Brianna Schofield University of California-Berkeley School of  Law: Tremendous diversity in how service providers interact w/the notice system.  DMCA classic: intake manageable by hand. Varying degrees of quality of notices. 
 
Another group gets a large influx of notices; not the predominant thing for most ISPs.  DMCA-auto: reacted to influx by developing automated systems to process. DMCA-Plus: beyond the requirements of the statutes.  Any reaction should take account of this diversity.
 
Claggett: could you do percentages?
 
Schofield: we surveyed a small #, but tried to have a representative sample of broader ecosystem.
 
DMCA classic is dominant by far.
 
Claggett: does this suggest that different policies need to be made depending on type of ISP?
 
Schofield: Instead, system is self-managing and reacting to needs of platforms. Hard to draw lines from policymakers’ perspective to arbitrarily putting any of them in buckets. Some service providers you might expect to get lots of notices are actually service providers that aren’t getting that many notices and don’t need automated processing.
 
Kevin Rupy USTelecom: Broadband service providers, large, small and everything in between.  Small rural providers with a couple thousand lines to “traditional” huge broadband providers like Verizon. Generally speaking, the safe harbor provisions are functioning as intended. However, we do have issues w/getting §512(c) notices when acting as mere conduit under §512(a); clear under statute and caselaw that those notices aren’t valid.  Multiple notices don’t render someone a repeat infringer; we believe that termination under appropriate circumstances should be narrowly construed and subject to some type of judicial oversight. Congress, this administration, and our main regulator the FCC view broadband employment/adoption as a principal goal. So termination needs to be for real reasons.
 
JC: if someone doesn’t pay their bill, do they get service terminated?
 
Rupy: Yes.
 
JC: so you do terminate subscribers for other reasons.
 
Rupy: Yes.
 
JC: Courts have said you don’t need a judicial determination to be a repeat infringer, at least some courts. Assuming that’s the law, how would a © owner communicate w/you about infringements repeating on your site?
 
Rupy: That’s happening in the marketplace, w/voluntary mechanisms like CCI: content community and broadband providers identify and address repeat infringement, or alleged repeat infringement.
 
Claggett: does it end in termination?
 
Rupy: depends on the nature of the escalation, it can.
 
JC: Your companies have agreed to take in notices and consider them voluntarily, but you don’t want the notices served under §512(c).
 
Rupy: if it relates to hosting, which we do, sure.  Where you have challenges is §512(a) conduit—there’s nothing for us to take down or remove!
 
JC: if someone becomes aware of P2P infringement on your network, how do they notify you? Do you accept those?
 
Rupy: Judicial due process; civil actions.
 
JC: but if a © owner wants to notify you, how would they do that outside of a judicial order?
 
Rupy: it would depend on the nature of the infringement.  If hosting that content, §512(c).
 
JC: instead of sending a notice, is there an acceptable alternative that would communicate the P2P infringement issue?
 
Rupy: we can talk about that, but as we view the framework currently, sending millions of notices to ISPs under §512(c) when they’re acting under (a) is not how the DMCA is set up. That’s one reason ISPs are working w/the content community for a voluntary framework for addressing repeat infringement.
 
Claggett: you’re saying a notice is just an allegation of infringement.  In your view, absent adjudication of infringement, would you not consider it unproven allegation?
 
Rupy: yes.
 
Claggett: These millions of notices: what is causing this phenomenon—just a difference of opinion on legal basis?  Is there any recourse in §512 to prevent that?
 
Rupy: That’s an improper use of §512.
 
Claggett: is this barred by §512 or can you use §512(f) or some other legal recourse?
 
Rupy: Can’t speak to that.
 
Claggett: Are specific stakeholders sending them?
 
Rupy: generally speaking, member companies have seen an increase over the last few years.  Different entities behind that; addressed in comments at broader level.
 
Claggett: ignore them?
 
Rupy: varies by provider and situation. Under some voluntary frameworks, you may see those notices forwarded on to the individual subscriber.  Even with the forwarding, it’s an allegation of infringement.
 
Jennifer Pariser Motion Picture Association of America: 512 is best for ISPs of every size. Google can automate; small ISPs don’t spend much; it’s a cost of doing business whether large or not. That’s where the imbalance comes: relatively manageable cost of doing business v. creation side is being killed by piracy and dealing w/great burdens from §512 to little effect.
 
In terms of §512(a): CCI is not a voluntary organization; it’s a contractually based organization with the MPAA, RIAA, and five largest ISPs, but only those 5.  Others are invited to be members but have chosen not to be.  Minimal effect of that program is only for subscribers of those services; only deals w/P2P piracy; doesn’t end w/termination.
 
Marcie Kaufman Ithaka/Artstor: we are stewards of content licensed to us and provide it to libraries etc. Have a freemium access model for users.  Sit between ISPs and users. We have to issue our own takedown notices. But we have SharedShelf ISP service model—content management for digital collections for institutions, integrated w/licensed materials. They do their own screening of IP rights, and are generally cautious b/c they’re academic institutions. Fills an important niche allowing teaching materials to be integrated. 
 
Artstor couldn’t have created this w/o DMCA safe harbor.  We don’t monitor it—it’s password protected.  Needs to be consideration of the service provider, size, etc. in any changes under consideration.  We are the innovation the safe harbors meant to protect and changes would really affect innovations like us.
 
Andrew Bridges Fenwick & West LLP: DMCA works well for ISPs; nobody’s happy with it, and that’s the sign of a good law.  Four points of concern, three of which can be solved by education. (1) Broad misunderstanding that the safe harbor has replaced the substantive standards of © infringement, or that the DMCA gives © rights to creators. It simply limits remedies. Before you reach safe harbor, you have to be otherwise liable.  Rampant confusion about that.  (2) Rampant confusion about role of §512(a) conduit provider and notices to them using §512(c).  (3) A number of abuses; DMCA used by business model itself, a monetization strategy and litigation strategy, rather than a framework to encourage collaboration.  (4) Needs litigation to fix: pervasive problem of extreme statutory damages.  Once that’s done, many other complex questions become easier to solve.
Kerry Sheehan Public Knowledge: Tempting to see future of §512 as being solely about ISPs and content creators, but also about internet users as a whole—increased risk of liability means increased risk of takedown of protected expression online.  [And they are content creators too!]
 
Jay Rosenthal ESL Music/ESL Music Publishing: Contrasting those who get the notices and those who send them. It’s great that it’s ok on the ISP side. On the content owner side, it’s an incredible drain, and smaller © owners have stopped using it.  Burden on ISPs, but there’s got to be a shift in this burden in policing the internet.  If we both feel pain, that would be the right balance.  But since they’re saying it’s working relatively well and we’re saying it’s not, that should push © policy towards us.
 
Damon DiMarco Author [Authors’ Guild]: One of only two independent © holders here. [I beg to disagree: what are AO3 users, chopped liver? I’ve sent takedown notices.]  Written five books. Google alert.  Forty alerts in April: pirated instances.  DMCA is toothless.  I spend the majority of my time running after people stealing from me.  Panamanian companies, overseas companies.  I should have an attorney, but I can’t afford one.  Independent artist’s POV: complete debacle. Don’t have an answer, but you need to know what the ecosystem is like. Stratified against independent producers of content.
 
Claggett: Types of improper notices—targeting legitimate content that have negative effects on free expression.
 
Hillary Johnson Author: What DiMarco said.  Major book published 20 years—most infringed, plagiarized book of the 20th century.  Most egregious: a FB user who’s published my book on his FB page three times over in 10 years. Started by sending takedown notices to FB; took about three days to satisfy FB’s demands to prove that I was the author who owned the ©.  I live on Social Security and a $500/hour lawyer presented 75 instances of © infringement. FB told this guy to C&D; I never saw the letter; he stopped for two months, but w/a lot of profanity, anger, hostility.  It stopped for a while but since then he’s posting material that appeared elsewhere on other people’s websites.  He’s now plagiarizing instead of posting verbatim. My book has been utterly devalued; efforts to protect it have been futile.
 
JC: when you sent a takedown notice to FB, do they respond? What do they demand?
 
Johnson: 2011-2013 I was sending notices to FB. They sent a form back asking for proof.
 
Thomas Kennedy American Society of Media Photographers: Incredibly asymmetrical. No collaboration w/individuals.
 
Mickey Osterreicher National Press Photographers Association: Haves and have-nots have completely different view of how this is working.  Fair use has become a sword—courts have difficulty deciding fair use, so that’s problematic.
 
Marc Ostrow Law Offices of Marc D. Ostrow: Silent majority: represents a lot of individual songwriters and recording artists and small businesses. There are lots of people who never bother to send notices any more to the YT and FB of the world.  Very few counternotices are even served because the person whose content was taken down will just repost it quickly.  Statutory damages aren’t that high when you consider the cost of prosecuting litigation to trial. I’ve had people get angry at takedowns b/c they say they’re fans.
 
JC: in your view people don’t use counternotice process, just repost.
 
Ostrow: in my universe of small publishers, yes. The one counternotice I’m aware of involved a small classical publisher—a university concert posted the entire performance, and the composer didn’t like the performance, and posting an entire work w/o transformation is not fair use; they claimed it was fair use.
 
JC: did you sue?
 
Ostrow: No, b/c of cost/benefit analysis. 
 
JC: did the content come down?
 
Ostrow: yes, it did, but in other circumstances it’s just reposted.
 
Claggett: want to focus on improper notices that are improper for free speech.
 
Pariser: The Berkeley study doesn’t have anything to tell us about the internet ecosystem outside of Google. The study looked at 1800 so-called improper notices w/in the confines of a limited period of time. 30% of notices sent were improper, supposedly. 15% were some variety of technical mistake; mismatch b/t Usher’s music and movie House of Usher—not suppression of legitimate expression. That work shouldn’t have been there to begin w/.  Others pointed to more than 1 URL, making it difficult for Google to know; that’s a problem for Google but not a huge problem.  7% in supposed fair use bucket.  They are way overgenerous in what might be considered fair use—covers and ringtones are not fair use.  Even so, doesn’t say © owner lacked good faith basis.
 
Google’s own statistics say less than 1%.  We won’t say there’s never been an improper notice in the history of the world. Anecdotal. Microscopically small percentage.
 
Schofield: Hosts of all sizes talked about notices targeting noninfringing content. The problem there is that there’s a tendency to take down content that the service providers do no matter what. Others have a different risk tolerance profile, like WordPress.  Study 2: largely automated notices often targeting problematic sites.  Large rightsholders are focusing their efforts on big sites.  So we looked beyond large rightsholders targeting those sites. So we pulled notices sent to Google Image search. Tended to be individuals/small businesses. Different dynamic, with lots more issues of expression: blogs, message board threads.  15% targeting improper subject matter.  Help us study!
 
Claggett: could you go back and determine whether those procedurally bad ones were based on legitimate content?
 
A: it would be hard.
 
JC: which is more accurate, automated or human review?
 
A: depends on type of rights holder.  Image search: human attention wasn’t necessarily a panacea. Largely unsophisticated actors in the system. 
 
Claggett: pro vendors typically had less inaccuracies than the individual or small business owner notices—is that who they targeted or their understanding of the law?
 
A: depending on how you count, 70% of Google Image notices had problem; we set one particular person’s notices aside, leaving 37% with problems in image search, and that’s about sophistication.
 
Claggett: does that suggest in preventing abuse there needs to be a different solution depending on the type of notice sender if individuals and smaller businesses—would higher damages make a different amount of sense?
 
A: we recommend tailored solutions. By & large, for automation there are best practices about refining these systems to minimize errors. We’ve spoken with rights enforcement orgs that say there are mechanisms to refine algorithms to limit mismatch.  Some orgs weigh success on numbers, not on quality. Smaller senders: educational efforts could be targeted to see what the process is about.
 
Rebecca Prince, Becky Boop: I’ve been subject of improper notices based on my fair uses. First: DMCA strikes by competitors. I create video on YT, which is one of the few platforms that gives you AdSense revenue, mostly w/in the first few days. Competitors can lob four notices = account termination; no recourse to reinstate your account. If this is how you make a living, that’s very serious.  YT’s Content ID system: I have the opposite problem, where I submit a counternotification, and the company just sits and waits and then just submit a new notice; they can keep your content down for months, which has a chilling effect and changes the content you create.
 
Claggett: are they not complying w/putback?
 
Prince: material will be reinstated, but company will create a new followup claim. It’s automated and so no one checks to see it’s a second claim to keep your content down again.
 
JC: Content ID?
 
Prince: yes, segues into DMCA if you counternotify.
 
Jay Rosenthal: Cover versions of songs are not fair use as Berkeley study says.  Small percentages—stifling innovation argument should be made much more from the user side. © takes the idea/expression distinction into account. If original © owner wants to stop use of expression, they should be able to do so. © owner needs remedy to stop someone from using expression.
 
RT: [I’m pretty sure there’s at least one PD film about the House of Usher.] [For examples of problem notices, see the comments of: Engine, Github, Redbubble, Matthew Neco’s discussion of Docstoc, Yahoo!, Automattic, Siteground, SoundCloud, Internet Archive, Wikimedia.] Jon Penney’s study of blogger and twitter takedowns: substantial minority percentage of plausible fair uses, including criticism. 
 
This is situation not ISPs v. copyright owners or so-called “content creators.”  Our users are creators and © owners too; we assist them w/information about DMCA when someone is selling copies of their works on Amazon; I’ve sent takedown requests myself.  But they are mostly at risk of invalid takedowns.  Mostly for critical works.  When we ask for proof of ownership, we do so b/c there are people who pretend to be HBO [Game of Thrones], Warner [Harry Potter], and all sorts of others.  The things people complained about in the last panel are the sorts of educational measures people have just been advocating for here—when Google asks if it’s a picture of you, the reason they’re asking is to try to figure out if you really have a copyright claim! You can’t have it all.
 
Haves and have-nots: those exist on both sides!  DMCA notice from big copyright owner—people are scared to counternotify b/c if they consult a lawyer the lawyer has to tell them about statutory damages.  One woman’s videos have been shown in multiple curated shows at museums; still thought long and hard about counternotice.  Small creators do often decline to interact w/the legal system.  But please recognize that if you accept the testimony that lots of people don’t bother to contest an abuse, that’s true for people who receive notices too!
 
Counternotices: Digital Media Ass’n, submitted a study whose questions assumed s512 wasn’t strong enough.  Has an interesting non-statistical survey, 9% of those experienced counternotices and also experienced invalid claims against them.
 
Michael Weinberg: I personally review every takedown. We get plenty of facially legit requests. But we get takedown notices for all sorts of stuff. The internet is big and people use a takedown option for everything. You may code something as a technical error—weren’t willing to assert they were the © owner—but what that means is that they didn’t actually have an ownership interest.  If we try to structure the complaint on the front end to avoid bad takedowns—taking the interest of creators seriously both for senders and for recipients—we have to balance burden on rightsholders sending notices and ability to avoid burdening our creatives.  We get people who send takedowns based on every use of a word they claim to own. 
 
Charlyn Zlotnik Photographer: 5 years ago I sent notices to ISPs and didn’t receive replies.  YT responded w/removals. Lately I’ve been going directly to people posting on Flickr or Pinterest; it seems like the burden was always on photographers but it’s out of hand now. There should be notices in real simple English—if you scanned something from a book, don’t post it. All my older work in books is being scanned in hi-res.
 
Perry Bashkoff WEA Digital & Revenue Development: Practicality of utilizing toolsets given to content owners: encourage decisionmakers to sit w/a content owner.  Tools are good enough; manpower hours to protect works of artists and talent is simply not manageable.  Practical reality of discussion: YT Content ID owner #1—seeing what it means to receive a counternotice saying “I bought this CD,” or “get lost”—we get that all the time.
 
Sheehan: No matter whether you characterize these as intentional or just mistakes, you get suppression of speech.  Every Single Word video series: edits films down to every single word spoken by a character of color. Excellent critique of racism in Hollywood; they received a DMCA takedown. What happens when content is taken down w/out a large public uproar?  Rightsholder rescinded a request.
 
JC: There is a counternotification procedure.
 
Sheehan: intimidating for users to swear—imbalance of responsibility b/t notice senders and counternotice senders.  SCt has made very clear that fair use includes expression, BTW.
 
JC: an unfair takedown notice—there is a procedure available under the law.
 
Sheehan: it isn’t working—so maybe we need to refine notices and make counternotification easier.
 
JC: education?  PK could do more to educate posters.
 
Sheehan: could also do more to educate rightsholders about proper notice. There are things to do to equalize burdens on rightsholders. Not simply an educational issue but intimidation factor of having to swear under penalty of perjury.
 
Schofield: if you extrapolate across the notices, that is millions of improper notices.
 
Adrienne Fields, Artists Rights Standard: Incredibly burdensome. Can’t reach a fraction of infringing uses on the internet. Notices that I file are rejected for no apparent reason. If they want to take on that role, they should be responsible for the infringing activity. Incentivized not to remove b/c they like having traffic from works by big name artists.  Rejected where the artist’s name was included on the video and thirty works appeared in the video. [All I want here is some recognition of the parallels b/t the “why don’t you sue if they’re not complying w/the DMCA?” question and the “why don’t you counternotify if the notice is abusive?” question.]  When we send notice to a conduit, it’s b/c we can’t find them any other way—they use proxy services.
 
Michael Housley, Viacom: Even if we had all the notices perfect and always sent them, piracy is growing. We need more than the notice black hole.  Filtering, other tech available now that can be used not just for antipiracy but for all sorts of content identification. [That’s not terrifying from a civil liberties perspective at all.]

No comments: