IPSC: Copyright Enforcement

Thomas H. Rousse, Northwestern Pritzker School of Law, Open Licensing, Hidden Costs: Survey Experiment Insights On Creative Commons and Copyright Infringement

Pragmatist; experience as journalist with taking a photo, licensing it via CC, and seeing it widely reused without credit was part of the impetus. ShareAlike licensing is often misused. 3Ls had difficulty finding right license. Generating logo for CC license was busted on Chrome and another browser; it always said in rich text that it was just CC-BY no matter what you chose (he agitated and finally it seems like that will be fixed).

History of CC; decided not to use warranty-like system. Uneasy relationship b/t ideological goals of changing from permission-based society to more open, less transaction-cost based permission system. Copyright registration (for US works) needed to enforce. Version 4.0 infringement mulligan—allows cure w/in 30 days; most people won’t know this happened and 30 days of use is often the entire value of the use.

Surveyed 1299 US adults w/quotas for gender, race, ethnicity, age. Collected demographic info including political preferences. Asked about logo/brand recognition. Also scenarios: Control and two treatments given.

21% self-reported recognition of logo. But when you ask, only 7% correctly identified Creative Commons; Closed Captioning, Comedy Central, Chanel, Copyright were common mistaken answers. Also not much recognition of Creative Commons (nearly 60% were unfamiliar), though it does better than Mozilla.

Control: maximalist © notice w/author, year, all rights reserved

Treatment A: Abbreviated CC marking (BY)

Treatment B: Training and verbose CC marking (attribution)—people had to read for at least 10 seconds before proceeding.

7 vignettes: public domain (noninfringing gov’t work), personal use, commercial advertising use, large scale willful copying (selling lots of posters for profit), educational use, Wikipedia licensing error, and filesharing—the last 3 actual scenarios that have had controversies/litigation. 

Asked about likelihood of legal consequences. Asked if the reuser was sued, what consequences would they likely face, then asked what consequences should they face.

When you don’t tell people what the public domain is, they think it’s infringing, but when you explain a little they don’t. Commercial ad use: educating people on what CC is does change how they perceive risks. Surprising amount of support for finding various things done in breach of license to be infringing.

Conclusions: Respondents were slightly more supportive of CC licensors than regular © holders. [To me suggests a “fair is fair” mentality—by being reasonable about your rights, you’re entitled to more consideration than if you were being unreasonable.]

RT: Super interesting! Question about ShareAlike: when you use a CC ShareAlike photo to illustrate a story, is the accompanying story really governed by ShareAlike? I wouldn’t have considered the story an adaptation of the photo, and the story is definitely not a copy.

A: disagrees—the whole website should have been ShareAlike if they use one ShareAlike photo. [I think that’s completely wrong, but the fact that two © lawyers can disagree about that might be relevant to your argument. FWIW a Quora user reports that CC agrees with me.]

Q: what should we be doing differently?

A: Spend more time educating people about what CC means. But that does tend to increase their support for enforcement against violators. [Sounds fine to me.] The CC trolls are also doing some education! “Freely shared” as an assumption about CC is not right.

Stefan Bechtold, ETH Zurich, Out of Tune? Investigating YouTube’s Copyright Enforcement

Focus on music and Content ID. At least 50% of videos on YT include music, and almost all most popular videos do. System creates a “fingerprint” & claiming owner can decide whether to block or monetize. Mostly they monetize.

Scholarly concern: replace © law, affect/eradicate fair use? Including me.

Comprehensively audit Content ID: Uploaded more than 10,000 videos w/music either © protected or not, focusing on copyright term and on rights in compositions rather than sound recordings. All uploaded videos were set to private. Record Content ID’s response. For example, Content ID can’t distinguish b/t different recordings/performances by Bach. Classical compositions clearly out of ©; we also used snippets from Spotify where there was © in composition and in song recording, using classical and pop music. 793 recordings, 790 compositions, 49 composers.

45% of works on public domain compositions w/CC recordings were flagged as ©-infringing. One driving force: © trolls. LatinAutorPerf makes 3x as many claims as the next most frequent claimant, Universal. Users may not complain b/c it’s just revenue sharing; so the troll makes a lot of money claiming to own the compositions.

What about large v. small labels? 4000 tracks from Spotify stratified for popularity and publisher status. 45.4% were unclaimed, 54.6% claimed. Four players behind over half of all claims. Major labels enforce at 37.5% higher than smaller. Content ID is only available to © owners w/most complex © management needs. Smaller players can use takedowns or pay third-party distributors, but those usually take 20-30% of revenue.

Brussels Effect: exploiting different © terms for compositions in the US v. EU. Some compositions therefore out of © in US due to 95-year term but in © in EU. In 60% of cases, YT flags them globally, even when they are in the public domain in the US.

Findings: tech works well, but shortcomings of institutional arrangement governing system. 45% overenforcement for public domain works, often claimed by © trolls, and 45% underenforcement for smaller players; Brussels effect of longer European term.

Also finding that probing digital platforms w/algorithmic auditing tools is a promising approach to understand and optimize them.

Q: is the problem really about rightsholders in composition specifically? What are they uploading? It has to be a recording. Content ID can’t really just have the composition itself qua composition. It’s always analyzing a phonorecord.

A: correct. Fingerprinting tech tries to be robust and derive the underlying composition, but can it?

RT: Love it. One Q: any chance these weren’t real small labels? Spotify has a lot of wholly Spotify-owned music. It might be indifferent to the presence of the same music on YT b/c no one ever seeks that music out; Spotify just stuffs those into playlists so it can keep more revenue for itself. So there might be less true underenforcement than you thought.

Q: I think a © troll is overenforcing real rights. This seems like © fraudsters, not trolls.

Beebe: is there further YT could go?

A: given it’s been around for a long time and large players have been involved for 2 decades, this can’t be totally surprising. Perhaps it’s just the evolved equilibrium. Maybe it’s cost-prohibitive to do more things. But the more we move into an automated decision/compliance mechanism, we’re very far from near-perfect accuracy and we need to recognize that.

Sidne Gard & Elizabeth Townsend Gard, School of the Art Institute of Chicago; Tulane U. Law School The Economy of Cultural Accountability, or the Economic Role of Moral Rights in the YouTube Age

Lost artists: Internet folktales based on photos etc. (Slenderman, Cooper Family Falling Body Photo—no one really knows who made it, how it was taken, how the story got attached to it, etc.) Turns out the Cooper photo was art made by Richard Ramsdell. YT video investigated the story, found Ramsdell, and convinced him to put the other images of the series online as well as selling merch. An orphan work was reunited with its author, changing its trajectory. VARA wouldn’t apply for attribution, but shows how attribution is tied to economics.

Similar story: Disney Channel theme song, used for over two decades, but identity of creator unknown until YT documentary found Alex Lasarenko, a composer and classical pianist, through intense research, interviews, and use of archiving work by third parties of things like commercials on the Disney channel. He passed away two years before being found, but 8 million people listened to his work in the documentary.

Lessons: WFH doesn’t mean there isn’t an artist behind the work. Finding an author means finding more of their work. Can © even be attached to a 4-note jingle? What is big enough to be a work? Is TM also relevant?

Other side of the coin is plagiarism: Big video essayist posted video about YTers plagiarizing other smaller creators’, journalists’ and writers’ work. The back half of a 4-hour video is entirely about James Somerton, who did video essays about queer theory; he was also taking 10,000 words in hour-and-a-half video from 18 different authors who went uncredited or poorly credited. He’d credit the major players of queer theory, but not the more unknown ones. Effectively ended his career.

Mob mentality risks, but also restored attribution.

Minecraft Redstone Prison Door controversy: you can mine redstone to make complicated machines w/electricity. People get very into redstone creations; make tutorials about things they design. But larger creators may use redstone builds they find online with credit (fine) or without (not). Lack of credit has economic impact because credit is how they grow their channels and revenue.

YouTube doesn’t have a way to rectify false attribution—Content ID doesn’t work to do that.

RT: some of these things might have infringed © and some not. Is this relevant to whether attribution can ever be legally managed or has to be managed on a community basis?

Q: how much of this is US-specific since we lack moral rights?

Q: what about community notes/annotations as a model? YT comments are not known for being the most friendly place, but is there a building block there?