You know, for a judge who loves the First Amendment and free speech so much when it comes to the right of publicity and nominative fair use, Judge Kozinski can seem oddly unconcerned about it in other areas of IP, like dilution, secondary liability, and, here, copyright. It’s not just the substance of the decision—the injunction here came with a gag order preventing Google from disclosing the mandated takedown until the opinion issued. The injunction was subsequently narrowed a bit, after it was made public; now anyone who edits Garcia out can post the video at issue.
Eric Goldman’s commentary is sharp but not unwarranted; the IP professors went at this case hammer and tongs for a while, with almost nobody happy with the injunctive remedy and most though not all displeased with the authorship analysis.
The real problem is that the 9
thCircuit (like other circuits) screwed up joint authorship with weird intent requirements, themselves apparently motivated by the non-statutorily-required rule that all coauthors receive equal shares. This had the further result that when Kozinski encounters a situation where the actor seems to have some equitable claims, the only apparent solution seems to be to give her a separate right in a separate “work,” making the doctrine a ton worse off.
Garcia v. Google, Inc., No. 12-57302 (9th Cir. Feb. 26, 2014)
The facts are probably familiar: Garcia was tricked into appearing in a film that she thought was “Desert Warrior” but turned into “Innocence of Muslims,” and because of dubbing her character appeared to utter one of the more offensive lines. Garcia was only given the four pages of the script in which her character appeared and paid approximately $500 for three and a half days of filming. As a result, Garcia started receiving death threats. After filing futile, §230-barred state law claims against Google, she registered a copyright in her performance (or the AV work featuring her performance, as it says on the registration—the nature of the claim is really unclear, even though Kozinski mocks the district court for saying just that) and sued when Google refused to honor her takedown notices.
Kozinski defined Garcia’s claim as one to own a copyright in her performance within the film as independently copyrightable. She’s not a joint author of the film under 9thCircuit precedent. The dissent thinks this claim is best analyzed as a joint work claim and that Garcia isn’t a joint author; the majority disagrees because there was no intention to create a joint work. But that doesn’t really get to the dissent’s point: if the work at issue is indeed one work rather than a bunch of subworks, then Garcia can only be an author if she’s an author of the work. And the only way to do that is to be a joint author, which under 9th Circuit precedent (quite plausibly wrong, but this is not the way to fix it) she can’t be. (What is the right definition of the work? I think it’s the film, which shouldn’t be divided into a bunch of independent subworks for a lot of doctrinal reasons, e.g., fair use, termination of transfers, etc., a position with which the Copyright Office expressed agreement last week when it denied Garcia’s copyright registration application.)
So, the majority continues: if Garcia made an independently copyrightable contribution to a joint work (I think this must mean “work,” because it would seem to be true whether the work is single-authored, joint, or WFH), she can retain a copyright interest in that contribution even if she’s not a joint author. All that’s required for copyright protection is a fixed, original work of authorship, and the “artistic contribution” here was fixed, so the only remaining question was whether it was sufficiently creative to be protectable. (Court’s footnote: whether the author of a dramatic performance “must personally fix his work in a tangible medium” was not here at issue, but can be raised on remand. “Personally” must be a reference to “by or under the authority of the author,” a requirement for a protectable fixation, and it is a rabbit hole I will not go down very far, except to say that the fraud claim and the “I authorized fixation” claim do not fit very well insofar as the fraud went to the sole reason Garcia authorized fixation.)
Google argued that Garcia’s contribution to the film wasn’t sufficient to make her an author, because Youssef wrote the dialogue she spoke, managed all aspects of the production and later dubbed over a portion of her scene. “But an actor does far more than speak words on a page; he must ‘live his part inwardly, and then . . . give to his experience an external embodiment.’” That embodiment (and thus, presumably, that authorship) “includes body language, facial expression and reactions to other actors and elements of a scene. Otherwise, ‘every shmuck . . . is an actor because everyone . . . knows how to read.’” (Hunh? Given what Kozinksi’s about to say next about the minimal creativity standard, isn’t this a holding that every shmuck is an actor?) A fixed performance is copyrightable if it evinces “some minimal degree of creativity . . . ‘no matter how crude, humble or obvious’ it might be,” per Feist. This was true of Garcia’s dubbed-over performance.
Aalmuhammed didn’t control because it was only about joint authorship. “It makes sense to impose heightened requirements on those who would leverage their individual contribution into ownership of a greater whole, but those requirements don’t apply to the copyrightability of all creative works, for which only a ‘minimal creative spark [is] required by the Copyright Act and the Constitution,’” Feist. (So Jefri Aalmuhammed was just a poor pleader. And now those who own a tiny bit can have more control than a joint author (who can’t prevent other joint authors from licensing/using the work.) Note too here how the question of what a “creative work” is has been carefully elided.) Contrary to the dissent, the majority continued, it wasn’t reading authorship out of the Copyright Act. (Only “work.”) The creator of copyrightable expression is an author, which is why Sinéad O’Connor has a copyright in her performance of “Nothing Compares 2 U” even though the song was written by Prince.
Now the majority turned to the right of publicity cases, which are a frolic and detour of their own: because courts have analyzed the right of publicity’s conflict with copyright as a question of §301 preemption rather than conflict preemption (which would be more sensible), and because the 9th Circuit (against Kozinski’s good advice) has expanded the right of publicity a ton, courts have previously drawn a rather evanescent distinction between an instance of singing (which can be fixed) and a voice (which, supposedly, cannot be fixed, even though that makes no sense whatsoever, and this distinction was unnecessary, but that’s a separate fight that nonetheless makes matters here even more complicated). Kozinski says that the right of publicity cases weren’t copyright cases, which is not even true, since they were at least §301 cases, though they weren’t infringement cases. Anyway, Kozinski continues, Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988), was about the copyrightability of a performer’s voice—not her performance. “A performer’s voice is analogous to her image, which we’ve said ‘is not a work of authorship’ under the Copyright Act. Downing v. Abercrombie & Fitch, 265 F.3d 994, 1004 (9th Cir. 2001).
But that doesn’t answer the question of whether the artist’s creativity, expressed through her voice or image, is protected by copyright. Just because someone’s voice—its particular timber and quality—can’t be copyrighted, doesn’t mean that a performance made using that voice can never be protected. In fact, many vocal performances are copyrighted.
True! But, and this I think is at least consistent with Kozinski’s preferred position, this stance would seem to put §301 preemption back into play for many right of publicity cases involving, for example, pictures of celebrities used with the consent of the copyright owner/photographer: since an image made using a celebrity’s face can be copyrighted, that particular image is back within the subject matter of copyright, like the sound recording of a voice. (Although under Garciaauthorship is now an unholy mess there, if the celebrity posed in any way.)
Setting that all aside: Garcia’s performance was a derivative work of the script, and one she had implicit permission to make because Youssef hired her, gave her the script, and turned the camera on her. She doesn’t own an interest in the entire scene, only in her contribution. Even if that contribution is relatively minor, “it isn’t de minimis.” The majority declined to hold that every actor has a copyright in his/her performance in a movie (though of course, given the few seconds Garcia is on screen, it is in practice so holding). “[W]hile the matter is fairly debatable, Garcia is likely to prevail.” (In its petition for rehearing on the stay, Google makes much of this statement—as a procedural matter, if it’s fairly debatable, a prohibitory injunction is an unusual outcome, especially one that reverses the district court.)
You might be worried that “the numerous creative contributions that make up a film can quickly become entangled in an impenetrable thicket of copyright.” But don’t fret: contracts, work for hire, or implied licenses will make such problems “rare[].” (I wonder how many studios have confidential memoranda in preparation about their ability to prove the existence of written contracts with every Tom, Dick and Harriet who ever showed up in a film, dressed a set, or otherwise contributed to the movie.)
Here, anyhow, there’s no work for hire issue because Garcia wasn’t Youssef’s employee, and she didn’t agree to a WFH arrangement (because she contended, without contradiction on this record, that the document purporting to allocate her copyright interests was a forgery). An actor in a small role probably isn’t an employee, because she was hired for a specific task, worked for only three days, and received no health or other traditional employment benefits. The dissent is wrong that Youssef’s control made him an employer, because there was no evidence that he directed the film or that he “controlled the manner in which any part of the film—much less Garcia’s scene—was shot. In fact, Youssef has claimed only that he wrote the screenplay.” Also, there was nothing in the record to suggest that he was in the regular business of making films.
What about implied license? Nonexclusive licenses can be implied from conduct, specifically where a plaintiff creates a work at a defendant’s request and hands it over, and where a plaintiff’s contribution to a film or other work would otherwise be worthless or of “minimal value.” That was true here. Garcia did grant the filmmaker an implied license, and it must be construed broadly. Failure to meet an actor’s expectations isn’t enough to exceed the license, or the copyright owner in the film could be unduly constrained in editing/creating derivative works or barred from distributing it entirely. Broad licenses are necessary to avoid giving actors de facto authorial control over films. (But: If Garcia were a real copyright owner, that wouldn’t be a weird result to be avoided. See Stewart v. Abend.) Also, these implied licenses should be interpreted broadly to protect joint authorship rules, since most actors don’t qualify as joint authors and shouldn’t have more control over a film than a joint author.
But a broad implied license isn’t unlimited. “Garcia was told she’d be acting in an adventure film set in ancient Arabia. Were she now to complain that the film has a different title, that its historical depictions are inaccurate, that her scene is poorly edited or that the quality of the film isn’t as she’d imagined, she wouldn’t have a viable claim that her implied license had been exceeded. But the license Garcia granted Youssef wasn’t so broad as to cover the use of her performance in any project.” The problem isn’t that Innocence of Muslimsisn’t an Arabian adventure movie; it’s not intended to entertain at all. (The majority knows this … how? It’s trolling, which many people find quite entertaining, and even if those people are awful, I do not get this distinction.) “The film differs so radically from anything Garcia could have imagined when she was cast that it can’t possibly be authorized by any implied license she granted Youssef.”
We know this because he lied to Garcia, and she relied on this lie. “Youssef’s fraud alone is likely enough to void any agreement he had with Garcia.” (In which case this holding is by no means narrow. If a performer alleges that part of the promise was the film’s name, or control over editing, or anything else, and then the agreement is therefore void, then the broad scope of the invalid license becomes irrelevant.) Even if the lie didn’t invalidate the agreement, it was clear evidence that his use of her performance “exceeded the scope of the implied license and was, therefore, an unauthorized, infringing use.” (Ditto.)
Filmmakers will rarely exceed the bounds of actors’ broad implied licenses, but this was such a case. (If these are truly implied licenses, then how do we know their terms? Why wouldn’t their terms at least be deeply up for grabs?) Thus, Garcia was likely to succeed on the merits.
And she suffered irreparable harm. Copyright infringement isn’t presumed irreparable, but the damage to her reputation and death threats are. Her delay in suing for several months wasn’t dispositive. But she took legal action as soon as the film received worldwide attention and she began receiving death threats, which was enough. The threats to her are ongoing and serious, and she has to take safety precautions when traveling and moved to a new home and business location. Absent an injunction, she’d continue to suffer these security threats.
Garcia also needed to show a causal connection between the harm and the conduct she sought to enjoin—the infringement of her copyright. She did, because the unauthorized inclusion of her performance in Innocence of Muslims led to the threats. Removing the film from YouTube would help disassociate her from the film’s anti-Islamic message. (How? How would it do so more than suing already did?) Google didn’t show that the film was so widespread that its removal from YouTube would have no effect (plus the court sua sponte ordered removal from every Google property and “all reasonable steps” to prevent reuploads, beyond Garcia’s requested relief, so there!). “Taking down the film from YouTube will remove it from a prominent online platform—the platform on which it was first displayed—and will curb the harms of which Garcia complains.” The publicity generated by the lawsuit was a necessary product of her attempt to protect herself, and not the source of harm to her. “To the extent the irreparable harm inquiry is at all a close question, we think it best to err on the side of life.”
The balance of equities also favored Garcia, because Youssef lied to her and Google has no equitable interest in preserving a use that she found abhorrent and that lead to death threats. Google’s argument that there was a public interest in preserving a record of The Innocence of Muslimsfailed because the First Amendment doesn’t protect copyright infringement. (I should note that Eldred and Harper & Rowdo presume that there is someone who wantsthe work to be disseminated, but to get paid for it; that seems to be an important difference.) The public interest therefore favored Garcia.
Judge Smith dissented.
First, the dissent said, a mandatory injunction—one changing the status quo—should be subject to a higher degree of scrutiny, requiring extreme caution. In combination with the standard of review, which is abuse of discretion, reversal was therefore not warranted.
Garcia didn’t have a copyright interest in her performance, because her performance was not a work, she was not an author, and her performance was too personal to be fixed. Originality is not enough for copyrightability/authorship.
The list of exemplary works in the Copyright Act are significantly different from an actor’s individual performance in a film, which makes the claim start out dubiously. (This list is nonexclusive, so the dissent rightly points to the dissimilarity, not just the failure to make the list. Pam Samuelson has a great draft paper about how to judge copyright suitability for cases like this, or gardens, or DNA sequences.) §101 also differentiates a work from the performance of the work. “While Congress distinguishes the performance from the work itself, the majority blurs this line.” An acting performance more resembles a §102-excluded “procedure” or “process” by which an original work is performed. “In sum, a motion picture is a work. A segment independently produced and then incorporated into a motion picture is also a work. However, the Copyright Act does not clearly place an acting performance within its sphere of copyrightable works.”
As for authorship, also a constitutional requirement, Aalmuhammed articulated general principles relevant to both single and joint authorship claims. Garcia’s claimed interest is best analyzed as a joint work (a claim which then fails), because she relied on Youssef’s script, equipment, and direction, and thus had the intent to merge their contributions into inseparable or interdependent parts of a unitary whole. An “author” is the person with creative control, and authorship requires more than a minimal creative or original contribution to the work—a principle the majority disregards. Garcia’s contribution was a lot less than Aalmuhammed’s. She “was not the originator of ideas or concepts. She simply acted out others’ ideas or script…. Indeed, it is difficult to understand how she can be considered an ‘inventive or master mind’ of her performance under these facts.”
Aalmuhammed is indeed relevant, because it articulated general principles of authorship, following Burrow-Giles, which isn’t a joint authorship case. (This is actually a revisiting of the “what is a ‘work’?” debate under another name.) The majority’s reliance on Feist was misplaced, because Feist was not about authorship.
The practical implications: Garcia’s minimal role in the film gave her a copyright, which will indeed create “an impenetrable thicket of copyright,” in contrast to Aalmuhammed, which, while debated in academic circles, “adopts a standard that promotes clarity in the motion picture industry.”
Fixation: if the relevant work is a “performance,” it’s not fixed. The right of publicity cases discuss “subject matter akin to an acting performance” and hold them to be unfixed. These are indeed “copyright cases,” coming from the same judges interpreting the same Copyright Act. Midler, for example, held that a voice isn’t copyrightable because its sounds aren’t fixed, and was “more personal than any work of authorship.” By contrast, other cases have found preemption where the defendant was using the actual copyrighted work in which the plaintiff had consented to perform. “Combined, these cases show that, just as the singing of a song is not copyrightable, while the entire song recording is copyrightable, the acting in a movie is not copyrightable, while the movie recording is copyrightable.” (The dissent distinguished cases in which an “independently authored clip” is used in a film, as in Effects Associates; its analysis was specific to an actor “acting out a script she did not write under the direction of someone else who provides all of the instruments, tools, and leadership.”)
Ultimately, a sound recording “involves many moving parts, including the tune, lyrics, instrumental musicians, vocalists, and a production team that edits and prepares the final song.” The result is copyrightable and the vocalist’s contribution is creative, but the vocalist’s singing of the song is not independently copyrightable. An acting performance is similar.
Separately, the dissent was also sympathetic to the WFH argument. Under the multifactor test, none of the factors is determinative, but the hiring party’s control is the central inquiry. Garcia conceded that Youssef “managed all aspects of production,” controlling both the manner and means of making the film, including the scenes featuring Garcia. Youssef also “provided the instrumentalities and tools, dictated the filming location, decided when and how long Garcia worked, and was engaged in the business of film making at the time.” “While the majority may dispute which person was actually directing the film, it cannot overcome Garcia’s own admissions in her complaint that substantiate these facts; she was not in control.”
The majority erred in failing to defer to the district court on likely success. There was no abuse of discretion in finding that Garcia hadn’t clearly shown likely success on the merits.
On irreparable harm, too, the dissent would’ve found no abuse of discretion in holding that Garcia’s delay negated irreparable harm, even given Garcia’s explanation. “Were Garcia really trying to protect her purported copyright interest in her acting performance, one would expect her to have brought this action immediately after learning of the alleged infringing behavior. Considering ‘[t]he relevant harm is the harm that . . . occurs to the parties’ legal interests,’ Garcia has failed to explain her delay in terms of harm to her alleged copyright interest.” Also, if she valued her acting performance to the extent she now claims, she could’ve protected it by contract. Instead, she acted for three days for $500.
On the equities, the majority trashed Youssef, but Google wasn’t a party to that conduct and Youssef wasn’t a party to this appeal. As for the public interest, there’s no First Amendment interest in infringing copyright (citing Eldred). But this isn’t “copyright infringement per se,” by which the court seems to mean “infringement of a freestanding work.” Instead, Garcia may have a copyright interest in her acting performance, but issuing a preliminary injunction (against the entire work) may be an invalid prior restraint.
There is obviously much more to come. I will just address one issue raised by this holding: James Grimmelmann, as part of his deconstruction of the decision on Twitter, asked what would happen to statutory damages now. 17 USC § 504(c)(1) provides that “For the purposes of this subsection, all the parts of a compilation or derivative work constitute one work,” so I can’t see how you can get more than one award of statutory damages per film even if it’s now composed of a bunch of independently copyrightable performances, since the film is either a compilation or a derivative work of those performances (or they’re derivative works of the script, or something). However! Precisely because of the one-statutory-damages-award-per-work rule, would anyone seeking statutory damages for infringing a film have to implead any actor with a potentially copyrightable interest, since his/her rights could be prejudiced by an award to the main copyright plaintiff? Or maybe most actors won’t have timely registered—do they have to, if the film copyright owner timely registered? By analogy to newspaper articles, I would guess they have to register separately in order to be entitled to statutory damages. So they’re probably not entitled to claim statutory damages in the ordinary case, whatever the “ordinary” case is here. Ok, then, but what happens if the film copyright owner elects statutory damages? Is the actor entitled to claim actual damages, even though that does violence to the idea that there is an election rather than a double recovery for an infringement of a single work as defined by § 504? As Grimmelmann says, what this really shows is that inconsistencies reverberate throughout a system.
Because of the panel’s take-down order, “Innocence of Muslims” is no longer on YouTube. However, if the Court wishes to view the film, it is available at many other sites on the Internet. Google will not list the URLs in this public document, but would be happy to file a representative list under seal at the Court’s request.
The Copyright Office also got in a whack at the 9thCircuit rule that application, not registration, is a prerequisite to suit: Citing 17 U.S.C. § 411(a), it wrote that “Congress expressly envisioned that registration decisions by the Register of Copyrights would precede adjudication in the courts” so that the Office can intervene to defend that decision. When a lawsuit comes first, “the Register’s statutory right to intervene in an action instituted pursuant to a refusal to register is nullified.” Which raises the question: will the Register nonetheless (move to) intervene?
Rebecca Tushnet 
