WIPIP: In Memoriam and Fair Use

A Few Words for a Lost Friend: Tribute to Dmitry Karshtedt (Bob Brauneis, Mark Lemley, Jake Sherkow)

Closing Plenary Session: Fair use

Robert Brauneis, Copyright Transactions in the Shadow of Fair Use

Suppose a work does not infringe another work because and only because it’s been ruled a fair use. Does the status of that work or copy limited subsequent uses of it or transactions about it? Yes, in some circumstances, and looking at that tells us some things about them.

Fair uses tend to divide into buckets: justified by new work; justified by project. Cites me and others on what I call content transformativeness and purpose transformativeness.

New work: Derivative work or embedding work: Cambpell v. Acuff Rose, or quotation of Patton’s Principles in Patton on Leadership. Use is justified by context of being placed in new work. Corrolary: fair use can be determined when new work is fixed in final form; determination won’t change after that; work can be performed or reporduced as a whole in any quantity, but reproduction or performance of part might remove the justifying context and be unfair: example—just the riff of 2 Live Crew’s Pretty Woman. The parodic context has to remain for the work to remain a fair use. Could not use excerpt of hook from 2 Live Crew for ad w/o permission of Acuff-Rose. Example 2: Axelrod book on General Patton quotes Williamson on Patton: fair use b/c it uses in the same way as any scholar does—but selling items just printed with that quote wouldn’t be fair.

Fair use justified by project: multiple copies for classroom use, time shifting vidos, visual search thumbnails, full text search engine, intermediate copying for reverse engineering.

Final determination of fair use is not possible when a copy is made. If the project is fine, it’s fine to repeat the project with additional works.  

Fair use and first sale: in tension? If you see fair use as a statutory license to use the old work under conditions that justify the use, then some of those conditions are about use of copy.  Does that mean that a videotape made for private timeshifting purposes can’t be sold or given to someone else, despite §109 first sale? Or does it mean that a videotape for time-shifting was lawfully made under this title?

Current thinking: neither. Adding a transaction changes the project so we have to assess whether this new project is fair use. If I videotape a show, watch it, then sell on eBay, hard to see how that project as a whole is fair use, particularly w/erasable, usable materials.

Google Books: Never displays full text in search results—fair use. HathiTrust uses those copies to provide print-disabled patrons w/accessible copies—requires different analysis, but also a fair use.

Not suggesting that “work” and “project” are exhaustive categories. Cariou v. Prince is work plus embodiment. Prince’s large, one of a kind originals didn’t usurp Cariou’s market. That should constrain use and licensing of the works if market separation into strata by value matters. Cariou is now selling prints from Yes Rasta for $1100 and Prince is selling exhibition posters for under $700—thinks it’s much less likely to be fair use.

Jeanne Fromer: Related question: when can you take a fair use decision and extrapolate from it.

A: comes up in Grokster concurrences w/battling characterizations of Sony 30 years later.

Q: Is Prince selling a work or a brand? That might make a difference to whether they compete.

Ochoa: Warhol may upset your argument, since there were suggestions that this was a content/work-based case, some uses might be fair and others might not be. Relatively separate in case law until now.

A: Cariou already does that and provides background for Warhol. Dct considered it an all new work. Licensed for use in magazines and don’t care; still fair use. Cariou is the case that might justify hybridizing.

Q: Third hypo: the videotape was made for time shifting years ago and is now a historical artifact. Is that still part of the original project?

Jim Gibson: consider 103 interactions too. Does 2 Live Crew have some © interest in the riff even if the rest that made it fair use/©able has disappeared?

Lunney: points out that Campbell was just a remand.

Wendy Gordon: Compare to negotiated direct license to do a cover (thus no requirement of consent to create a separate protectable derivative work)—does licensee of the cover version need permission of both?

Rosenblatt: how we think about that might be different in music sampling than other derivative works. But how does this map on interim copying? If it’s only inside the computer, and you’re making a fair use, then the interim copying is fair use. Or: the interim copy is something no one ever sees, so it’s fair no matter what; it’s the later thing that’s an infringement or not. Does your approach give us information about that?

A: if you make the interim copy for a particular purpose, but do something after that, think about the project as a whole. [This will also have statute of limitations implications.]

Buccafusco: why not answer it as a matter of copying in fact? I didn’t copy the P’s work, I copied my own. If © attaches to works, why wouldn’t that be true?

A: copying a derivative work is also copying the underlying work in most of © law [except damages].

Silbey: statute refers to fair use of a work; perhaps every use has to be on its own terms.

A: but if creating Pretty Woman is a fair use, it should be done on a work level (at least for uses as a whole) w/o having to think about use on a more granular level; otherwise we get silliness like trying to decide whether you can distribute the recording in Florida versus New York versus California.

Mehtab Khan, Recalibrating Fair Use

How to think about mass copying to create image and text datasets (facial recognition, art generators, ChatGPT), including beyond content generation.

How do we address the data collection stages of dataset development, and how should we assess the public interest?

Stages of dataset development: problem formulation (is this an image recognition dataset?); data collection (does this involve any injury? It’s where the copying takes place); data annotation (subjective decisions: creating meanings attached to data); model implementation.

Data annotation might be where transformative purpose is assessed.

Trends: wide allowance for fair use. But fair use could be a corrective measure where the datasets are used to create harmful tech or exacerbate bias. Need a framework that takes process of creation into account in the first place before assessing purpose. Consider whether the tool is purely private or available to public for end user good purposes. Intermediaries determine nature of use and amount used, and (?) effect on the market. How should we consider users’ commercial benefit (using art on sweatshirts) versus intermediaries’?

Glynn Lunney, Transforming Fair Use

Why don’t courts see that everything is fair use?

Goldsmith’s argument—even if creating them was fair use, licensing them for magazines wasn’t fair use. But 2 Live Crew sold CDs as did Orbison. 2d Circuit didn’t pay much attention to this, relying on the idea that if this is fair use, movies made from novels would be fair and overbroad fair use guts the derivative work right. That also seems a red herring to him. We need to figure out what limits, if any, exist on the derivative work right just as much as we need to figure out what limits, if any, exist on fair use. Nothing in © Act says what the relative size of those two slices must be, and 106 says subject to 107-etc. while 107 says “notwithstanding.” That suggests a relative hierarchy: fair use is the one that comes out ahead.

Why is novel-to-film the touchstone? Why provide that right? Political economy, historical path dependence—movie industry was new and easy to oppress. But here are rationales: (1) there will be only one film b/c there’s a natural monopoly/high fixed costs; (2) to ensure the book gets written; (3) to increase the book/movie producer surplus combined (closest to truth).

Is there only going to be one move? Abie’s Irish Rose/The Cohens and the Kellys. There was an authorized film version of both! In digital era, we can get multiple high quality versions of a character all the time; Elementary, BBC Sherlock Holmes, Robert Downey Jr. Sherlock Holmes. It’s not Highlander!

Is it a necessary incentive for novelists?  Empirically, no reason to think so. In music industry, collapse of revenue stream didn’t stop music creation and skewed demand for popular works also leads to backwards-bending labor supply curve: low revenue leads to more music. GRRM hasn’t released The Winds of Winter—is the delay b/c we didn’t give him enough $, or b/c we gave him too much? Robert Jordan’s Eye of World—kept extending the series until he died. George Lucas earned so much from merchandise—is that why we had to wait for 16 years after Return of the Jedi?

There is more profit when you have more exclusivity, so filmmakers probably prefer the broad derivative work right, but does it serve the progress of science in any useful way?

Even if you think novel-to-film is the right model, should we extend it to other situations? Even here, the analogy fails as to Goldsmith. Licensing in Goldsmith would be incentive-redistributing: taking $ from one author and give it to another. Not obvious why © would prefer one over the other; 2d Cir. characterizes her work as the primary work and his as secondary: why use these hierarchical terms? © has no policy preference between earlier and later works. GvO rejected the film analogy: O wasn’t going to expand from desktop to mobile; they’re different, noncompeting markets; G’s system is a complement and not a substitute. That’s of course equally true for books & movies: being an author doesn’t give you the ability to make a film.

Breaking the cycle: start to talk about how losing revenue will harm follow-on creator. The way to break the circularity is to look at empirics. We’ve gotten a little better at that over time—rapid rises/falls in revenue and their effects on creativity—they don’t seem to have any.

Rosen: Moral rights? [he doesn’t believe in them]

Khan: Berne requires them; the US says it complies with them largely through the economic rights.

A: [he doesn’t care] The derivative work right is limited by fair use in US law, and is incorporated into both of those treaties as far as the US is concerned—but we still need to know if it’s the right thing to do. Not convinced it’s the best thing to do for society to give her more money from Warhol.

Khan: we’d be saying she should have more control, not necessarily more money. [But control is now done with; we’re really talking about money, as is Goldsmith.]

Sheff: progressive taxation would deal with the backwards-bending supply curve. You could also fight the market structure that captures so much of the value before it ever reaches artists, which can also account from distorted incentives. Do we need fair use to do this, or can other reforms make the problem go away?

A: will take his chances persuading 5 Justices v trying to get progressive taxation passed; only handles part of the misfit b/t copyright and purpose.

Q: is it enough if my airbrushed photos have a different meaning/message to me than the untouched ones?

A: Advocating for Campbell’s standard, reasonably be perceived.

Betsy Rosenblatt, Putting the Fairness in Copyright Fair Use

Is fair use a social justice tool? When she says it is, gets response: often a tool for exploitation. Both of us are right, but how do we make my position more right? How do we shape fair use so that it is more favorable to social justice? criticism, commentary, talking back to dominant culture benefit from fair use; access to knowledge benefits social justice; but appropriation by dominant groups with impunity does not. We do intuit that the teen writing fan fiction is in a different place than Marvel; think about inherent, sometimes hidden hierarchies—appropriating medieval France is not the same thing as appropriating oppressed cultures.

Fair use sometimes favors underdogs, which gives hope for it as social justice tool. What would it take to help underdogs more?

Why do overdogs win? Courts are more likely to respond well to the transformativeness/recognize the transformative merits of work by famous people/people who are admired. Also, unlitigated cases and risk/uncertainty aversion: fair use is both predictable and uncertain, leading people w/more resources to be more bold about fair use, while others will be afraid of having to face litigation. And of course repeat players usually win. Fair use doctrine could better promote progress if it took party resources and cultural market share into account. Add relative power or resources to the list. We already treat hobbyists and students better than market participants; this would be consistent with market failure theories of fair use. Consistent w/theory behind reparations, as KJ Greene has suggested.

Covert socioeconomic engineering in © has been going on for generations, so why not make it overt?

Should the less privileged always win? No, but a little player’s use of a big player’s work is less likely to usurp the big player’s market than the reverse. And this won’t always be relevant—big fights big all the time, and little little.

Her test wouldn’t help Warhols and Richard Princes of world but would put them in context, and help artists who are just starting out; people who are learning to make derivative works and struggling to professionalize.

The haves may have invested time, money, and personality making something that the have-not just uses. We’d care if we thought this would deter haves from making stuff, but the marginal incentives for the already prosperous are limited. The marginal value of fair use to the have not is greater than the marginal value of not fair use is to the have. Jay-Z should have more rights to what he did earlier than what he did later.

Andrew Gilden: hierarchies should be connected to the purpose of fair use. A member of a marginalized group may not have a creative motivation for the use—difference b/t sampling b/c lack of resources to make own sounds and going to the Gagosian and making copies b/c I can.

Fromer: would this disrupt the predictability of looking at patterns in fair use cases? If you don’t know how power imbalances play into thinking, harder to predict, which is worse for people who can’t throw money at the problem.

A: we can’t use the Warhol story as precedent for the flipped version of the Warhol story, and wouldn’t be able to in her story.

Q: road to justice is unlikely to be paved with more landlords, so wouldn’t look to ©.

A: I want redistribution too, but I’m at a © conference.

Q: timing—fixed at creation of first work—Jay Z has more copyright interests now in his early stuff than his late stuff? Why isn’t it assessed at the time of the accused use? This goes to predictability too.

Alex Roberts: a lot of these conversations could be about TM too, as well as ROP. OMG Girlz v. OMG Dolls in which the court’s order says “you can’t talk about cultural appropriation in my courtroom,” and granted a mistrial when some deposition testimony got in calling it racial appropriation. Is that relevant to your story?

A: not to adding another factor, but in the larger sense we should care who the parties are and not declaring mistrials if cultural appropriation is mentioned.

Said: nobody wants to talk about race in these cases—worry about backlash/appellate resistance. If the parties are going to prove up disadvantage, doesn’t that depend on who has the better lawyering? The Pepe the Frog guy didn’t mean to make a hate symbol. One story is fans simply engaging with the meme; the other story is misappropriation.

Consider fee-shifting and compare to civil rights statutes—one of the reasons fair use cases are the way they are is that Google has money to burn on fighting until they win. One-off defendants often can’t, and neither can one-off plaintiffs.

Charles Duan: GvO’s fourth factor language about public benefit/downstream benefits might help you. Consider also what’s done extrajudicially through Google, Amazon. We don’t have the opportunity to litigate the sort of rule you want. So how to get those systems in the direction you suggest?