Oracle v Google (Part 2): Was it Fair Use?

Bringing Merger in Through the Side Door

This is part 2 in a series (of at least three posts). Part 1 is here.¹Do I need to disclose that I’m actually from Mountain View? I guess I just did. I remember when the Googleplex was Silicon Graphics and there wasn’t even a Shoreline Park or an “Amphitheatre”.

Regular readers of the IP Breakdown—both of you—will recall a semi-regular feature called Is it Fair Use?, which is a fast-paced game in which I give you a lot of facts and an analogous (and often binding) case, and let you decide whether the use of the copyrighted work was a fair use or not. This game has been sweeping the nation for about four and a half years, apparently in layer of negligible thickness.

My point in this feature is that close fair use determinations are unpredictable because they are based on—at least²The famous four fair-use factors are not exclusive.—four fact-intensive factors that often end up as a kind of cultural value judgment. While many fair-use determinations are pretty easy, the ones that end up being decided by a court are usually pretty hard.

We are Google. This is the Googleplex. The old rules don’t apply to us. Also, these cute Android statues rise at night and suck the innovation out of small companies. Photo by Runner1928.

Juries Are Scary.

The fair use issue in Oracle v. Google represented a kind of triple witching. First, it involved fair use. Second, it went to trial. Third, it went to a jury trial. Twice. Most fair-use determinations seem to be made by judges, on summary judgment or even Rule 12(b)(6)—which is very strange, since multi-factor, balancing, fact-intensive decisions are usually left to juries.³One argument against letting the question of fair use go to juries is that fair use is now a Constitutional inquiry—specifically free expression under the First Amendment—and such inquiries have traditionally left to judges as guardians of Constitutional rights. I did a quick search on a commercially-available database of caselaw and found only seven appellate decisions (reported and unreported) where the question of fair use was put before a jury (in a copyright case).

Juries generally terrify lawyers and everyone else. Lawyers are terrified because hardly any of them have ever been before a jury.⁴This is for two reasons. First, jury trials—and trials generally—are becoming rarer. Second, few clients trust juries with lawyers who lack jury-trial experience, which means jury-trial experience becomes more concentrated in just a few lawyers. Everyone else is terrified because “jury” means either one of two awful things: (1) serving on one, or (2) someone like your (or worse, your dumb neighbor) making a decision that had real consequences for other people. But for some reason, jury trials were considered so important to the Founders they enshrined them in the Seventh Amendment to the Constitution.⁵To save you a click, it’s because juries frequently ruled in favor of colonial defendants against the Crown (and its perceived allies), even where the law clearly favored the plaintiff. This is known as “jury nullification.” At the time, there was genuine concern juries would routinely nullify contracts, but that fear has mostly proved unfounded. Yet, they have in some respects more power than the judge. Their verdicts are difficult to overturn. And unlike judges, they don’t need to explain their reasoning.

Explaining an Enigma to the Enigmatic

Since juries aren’t knowledgeable about the relevant law, they need to be instructed in the law. For some reason, we instruct them in the worst possible way: by reading them 20, 30 or 50 infelicitous instructions at the end of proof, i.e., right after they’ve endured days of confusing, infelicitous testimony. It’s sort of important to get them right. One of the few ways to overturn a jury verdict is if a jury instruction was both highly relevant to the jury’s deliberations but seriously incorrect. The remedy? New trial. It’s hard enough to draft jury instructions so they are accurate and unbiased.⁶For settled areas of the law, there are “pattern” jury instructions, put together by committees of lawyers and judges, which are supposed to be accurate and unbiased. But even these will become dated or reflect bias. What if the law is unsettled? What if it’s amorphous or difficult to explain?

What if the law is fair use? There’s a challenge. Even if fair use were a stable body of law, it would be difficult to explain. And it is not stable. And there is little agreement about it. Coming up with a set of jury instructions that doesn’t piss someone off is probably impossible. If you’re unfamiliar with the wiles of fair use, you might start with my first Is it Fair Use? problem. It dates to 2011, so it’s already a bit out of date.

The jury instructions in Oracle v. Google may be found here. (The ones about copyright start with Instruction no. 21.) I’ve read through them several times, and, honestly, they seem good to me—or at least, they don’t seem reversibly poor.⁷Bonus points for not being boring pattern instructions; the judge really seems to be trying hard to explain the law to the jury. Keep a tab open—we’ll be returning to them.

“Hard Merger” Lite

There are four statutory/classic⁸The Copyright Act just “codifies” the same factors that date back to at least the 19th Century. fair-use factors that you’re supposed to consider:

1. How the copyrighted work was used (“transformative,” commercial, parody, etc.).
2. How expressive (read “creative”) the copyrighted work is.
3. How much was taken and how important it is to the copyrighted work.
4. The effect of the use on actual and potential (but not theoretical) markets for the copyrighted work.

I explicate them somewhat here, but for now, this is good enough.

Remember last time I talked about “hard merger”⁹As distinguished from “easy merger,” which was Google’s merger theory that attacked the very copyrightability of the Java API organization. “Easy merger” proposed that the organization merged with its (unprotectable) functionality because Java programmers had come to rely on the organization in order to efficiently use Java. The trial court accepted this theory but was reversed by the Federal Circuit because the theory looks at the wrong time: it looks at how things are now, instead of how thing were when the APIs were being organized back in the 1990’s. Merger is determined at the moment of creation, not later., which was a theory Google never asserted. “Hard merger” proposes that, when Java’s APIs were being organized, that organization was dictated not by “creative” choices (remember “creative” here is a very low bar), but by the need to be easy and efficient to use by later developers. I also thought Google had a good reason to avoid it: the proof was really hard (hence “hard” merger): Google’d have to go back to the 1990’s AND it would have to prove that no choice in organizing the APIs was creative (and there were a lot of choices).

But “hard merger” translates well into a fair use argument. When you’re saying that outside considerations determined a work, you’re also saying the work isn’t very creative, and that goes to factor no. 2. And the organization of a bunch of APIs isn’t very “creative,” as that term would be commonly understood, to begin with. The advantage for Google is that it doesn’t have to prove the organization was entirely devoid of creativity, only that it’s mostly uncreative. Even better, it can appeal to the jury’s common sense (or prejudice¹⁰One person’s common sense is another’s prejudice. It’s prejudicial to say that computer code is less creative than, say, a novel. It’s also common sense.) and not so much on technical testimony.

Little Used Factor No. 2 Comes Off the Bench to Score 40.

Let’s talk about factor no. 2 for a moment. This factor is probably the least-loved fair-use factor. When the last Supreme Court decision to analyze fair use, Campbell v. Acuff-Rose Music, Inc.¹¹The 2LiveCrew “Oh Pretty Woman” case., discussed it, it got a single paragraph, compared with ten for the first factor, seven for the third and seven for the fourth. It comes in for a good deal of criticism because it implicitly gives some copyrighted works stronger protections than others, because uses of more “creative” works are less likely to be found fair uses. This is contrary to one of the hoariest dictums in copyright law: courts don’t pass aesthetic judgment, and works are not excluded from protection just because they don’t seem artistic enough. But, apparently, in copyright law, as in Animal Farm, all works are created equal, but some are more equal than others.

Here is how the Supreme Court has described the second factor (citations omitted):

The second statutory factor, “the nature of the copy-righted work,” § 107(2), draws on Justice Story’s¹²This is a reference to the fair-use case that first cast the familiar four factors, Folsam v. Marsh, which dates to 1841. expression, the “value of the materials used.” Folsom v. Marsh, 9 F. Cas., at 348. This factor calls for recognition that some works are closer to the core of intended copyright protection than others, with the consequence that fair use is more difficult to establish when the former works are copied. See, e. g., Stewart v. Abend, 495 U. S., at 237–238 (contrasting fictional short story with factual works); Harper & Row, 471 U. S., at 563–564 (contrasting soon-to-be-published memoir with published speech); Sony, 464 U. S., at 455, n. 40 (contrasting motion pictures with news broadcasts); Feist, 499 U. S., at 348–351 (contrasting creative works with bare factual compilations); 3 M. Nimmer & D. Nimmer, Nimmer on Copyright § 13.05[A][2] (1993) (hereinafter Nimmer); Leval 1116¹³This refers to Judge Leval’s classic law review article that gave us “transformational use.”.

Here’s the core jury instruction from Oracle v. Google, describing the second factor:

28. The second statutory factor is the nature of the copyrighted work. This factor recognizes that traditional literary works are closer than informational works, such as instructional manuals, to the core of intended copyright protection. Creative writing and expression like at the very heart of copyright protection, so fair use is generally more difficult to establish for copying of traditional works than for copying of informational works. The focus of this statutory factor is on how close the used material is to the core values of copyright protection. The less the used material implicates the core values of copyright protection, the more viable will be fair use and vice versa.

This seems like a perfectly reasonable way to describe the second factor. The instructions actually go on through Instruction No. 32, putting the second factor into terms relevant to the facts of the case. I see nothing at all wrong with these instructions.

All Fair Use Factors Are Potentially Equal, but Some Are More Equal Than Others.

Florian Mueller of FOSSPatents, writing before the verdict, thinks these instructions are “flawed” because they “suggest that something functional (such as program code) gets weaker protection than something very creative.” But the second factor is all about giving some types of works less protection than others, and it’s completely consistent with Campbell that functional works would be on the less creative side of things. But you don’t have to take my word for it: you could take the Federal Circuit’s word for it, in the same opinion that reversed Google’s earlier victory (scroll to page 1375):

Because computer programs have both functional and expressive components, however, where the functional components are themselves unprotected (because, e.g., they are dictated by considerations of efficiency or other external factors, those elements should be afforded a lower degree of protection than more traditional literary works.

This is why you saw Oracle trying to emphasize the creativity of the organization, by comparing it to more common types of creative works, like painting. This effort was probably doomed: Oracle just ended up contrasting the organization with things the jury intuitively understood to be creative. Oracle also elicited testimony about how the organization is more of an art than a science—which is true enough—but it’s an art that dictated by a lot of considerations. The goal isn’t to come up with a pleasing organization, but an intuitive one, the “best” one. The results are objectively measurable, not a subjective matter of taste.

Google thus started out with a major structural advantage: it had one of the four factors in the bag, and didn’t have to work to hard at it. In practice, the second factor is almost always the weakest factor, usually an afterthought, while the first and fourth factors get most of the attention. But the law says—and the jury was instructed—to weigh the factors however the jurors liked. And I’m guessing the jurors weighed them about the same, just because that’s the natural thing to do. So all Google had to do was fight Oracle to a draw on the other factors.

Phantom Factor Four.

It would seem, however, that Oracle had its own built-in advantage: the fourth factor, which relates to markets. The story Oracle had to tell was easy. Once Android was released, it sucked up all of the marketshare for what might be called the non-iPhone OS market. If there was going to be a cheap alternative to the iPhone, Android was the only way. But that market was rightfully Oracle’s to exploit.

Based on the live-tweets and other commentary, it seems this story went wrong. It went wrong because of Savaje, a company that was trying to create its own Java-based mobile OS and that Oracle’s predecessor acquired for this reason. The effort failed. I think Oracle was trying to show how Android’s dominance foreclosed competitors, like Savaje. But I think it ended up showing the jury that creating a mobile OS was very hard, and that neither Oracle nor anyone else was really going to succeed.¹⁴It didn’t help that the Savaje phone looked horrible. Which was unfair—all phones looked horrible back then. Savaje folded its tent before iPhones were released.

If so, this was a blunder. Oracle wasn’t obligated to prove that it or anyone else could have produced an operating system as successful as Android. Oracle didn’t even need to prove that it had the capability of exploiting the market: it could license the copyright, after all. All it had to prove was that there was a market and Android sucked it all up. Nothing in the jury instructions obligated Oracle to prove more (start with Instruction No. 34). True, a story in which Android squashed a rival OS with real potential would have been more powerful—and maybe Oracle felt that, with the way the other factors were going, it needed to overcompensate—but it’s not a story that held up. Juries like stories, but they want them to be true.

The Tao of Factor Three: How Much and How Central of WHAT Exactly?

The third factor—amount and centrality of the taking—is fascinating because it could flip dramatically depending on how the jury defined the scope of Oracle’s “work.” If you define the “work” as just the organization of the API library and the resulting declaring code—which was all that was at issue—then it appears that Google took the whole thing. If, on the other hand, you define the “work” to be the entirety of Java, then it’s a trivial amount: the way you organize your files versus the content of those files. This is especially so if you see the declaring code as just a reflection of the organization.

What constitutes a “work” is a philosophical question that seemed intuitive to the drafters of the Copyright Act—there’s no definition—but has real world effects. Taking three pages from a six-page short story is worse than taking three pages from a 600-page novel. This makes intuitive sense: with three pages, your half-way to understanding the short story—you might feel you don’t really need to read the rest—but isn’t likely to tell you much about the novel, even if it were crucial to the novel.¹⁵But what if it “spoiled” some major plot point? But the practical result is the novel, though filled with much more creativity, gets less protection. And consider this: what if the novel were comprised of short chapters, averaging six pages, that were only loosely tied to each other and could almost stand on their own?¹⁶Perhaps music is a better example: what if you took one half of one section of a twelve-section orchestral piece, where each section can be—and often is—played its own?

Here, the jury seems to have defined the “work” as the APIs as a whole, not just the organization. This makes sense since Oracle doesn’t distribute or license just the organization and declaration code. That’s be like giving someone the way you organize your files but not the files themselves. But what if the APIs themselves were public domain¹⁷They are not. Google avoided infringement by re-creating the implementing code using a clean room. but the organization was not? In that hypothetical, Google would have taken 100% of the copyrighted work, since all that was copyrighted was the organization and the resulting declaring code.

Heavyweight Factor 1: Does Transformative Use Just Mean “Productive Use”?

The score so far: factor 2—heavily for fair use; factor 3—for fair use; factor 4—push or maybe slightly against fair use.¹⁸Again, you’re not supposed to use the four factors as a kind of scorecard, and the jury was warned against doing so, but I’m guessing the jury did. Can Oracle look to the heavyweight fair-use factor: factor 1, the character of the use.

According to the jury instructions (start with Instruction No. 23), there are two aspects to the first factor, which isn’t quite how I think of it. The first factor is whether and to what extent the use was “transformative,” a controversial concept that is nevertheless an established part of the law. The second factor is whether and to what extent the use was commercial.

Ironically, “transformative use” was developed by Judge Leval of the Second Circuit as a means to help regularize fair use. It received a firm stamp of approval from the Supreme Court. It quickly became as squishy and difficult to apply as fair use itself. Worse, it tended to overwhelm all of the other factors, so that fair use analysis sometimes became nothing more than transformative use analysis. I’ve complained about this trend often.¹⁹My message to smart people everywhere: stop trying to regularize fair use. You’re making it worse. Let’s just accept fair use for what it is: an unkempt mess that manages to hold down a steady job.

Still, transformative use is a useful concept. The jury instructions define as: “add[ing] something new, with a further purpose or different character, altering the first use with new expression, meaning, or message rather than merely superseding the objects of the original creation.” This language is taken almost directly from Campbell, so it’s on solid ground. Thus, recreating scenes from an actor’s movie roles in order to tell the actor’s biography is a transformative use. Even if the movie clips were identical, the new context changes the way we perceive them: not as part of a movie but as part of the actor’s life. But if we had played several of the clips back-to-back-to-back as part of retrospective of the actor’s career, the context, though just as different, doesn’t change the way we view the clips as much. We’d focus on the actor’s performance and not on the plot, but we’d still be interested in the artistry of the movie, as opposed to the actor’s life. Somehow that seems less transformative.

It is not easy to translate “transformative use” into computer software. Here, Google took the organization of the APIs and put it in service of an mobile operating system. On the one hand, the organization is serving the exact same purpose as it did with Java. On the other hand, the overall project was different and more useful. This, it seems to me, is more like the actor’s retrospective than like the biography, and is thus only somewhat transformative.

But the jury might have heard something else when the court was defining transformative use. It may have defined the term as a productive: the use of the underlying work to do something new with the work, to push the work further. To do something “cool” with the work. If so, the jury might well have found the use highly transformative.

The other consideration for the first factor is commercial use. Google could hardly claim otherwise.

Before we call this a push, notice how this factor gets entangled with the fourth, especially with the way Oracle was putting on its proof. The jury might have thought: if Google hadn’t just grabbed the API organization and run, we might not have any mobile operating system that competes with iPhone. We all benefited from Google’s indiscretion.²⁰This is almost exactly the logic of the TVEyes case. Savaje just showed how hard it was to make a mobile operating system and to get it adopted. I have criticized this type of thinking because it privileges the economic side of copyright over the creative side (when I think they should be carefully balanced), but perhaps this way of thinking is also intuitive—at least where “functional” works are concerned.

The Verdict Will Stick.

But let’s be conservative and call factor 1 a push. With factors 2 and 3 in favor of fair use—and heavily in the case of factor 2—and factor 4 a push or maybe slightly against fair use, we cannot call the jury’s verdict obviously wrong.

And “obviously wrong” is pretty much the standard for challenging a jury verdict, either in front of the judge, or on appeal. A jury verdict is overturned only if no reasonable jury could have issued the verdict it did. My torts professor used to jokingly describe the standard as “the jury was insane!” But insanity isn’t really accurate. The classic case would be where the plaintiff had to prove several elements to establish its claim and failed to present evidence as to one of the elements, but the other elements were fabulously well supported. The jury, persuaded by the mountain of evidence, finds for the plaintiff, but the court enters judgment for the defendant because of the hole in the plaintiff’s case.

In Oracle v. Google, that’s even less likely to happen because there were no “elements” to be proven, only “factors.” An element is something that’s either been proven or not. To take a copyright example, an element of infringement would be proving you own the copyright at issue. Your evidence of infringement might be devastating, but if you don’t present evidence of ownership²¹If you registered your copyright timely, this proof is simplicity itself: just show your registration., you lose. “Factors” are used when the law is more amorphous. They give us ways to structure or think about evidence, but they are none of them really required. Google wasn’t obligated to put on evidence about any one of the fair-use factors (though it would have been well-advised to). Failure to put on evidence about, say, effect on the market wouldn’t have been grounds for throwing out a jury verdict. Instead, the judge (or court of appeals) would have to find the evidence so thin, across all factors, that the jury really was nuts to find for Google. We’re not even in the same star system of that.

UPDATED (6-9-16): Apparently, while I was writing this, the court upheld the jury verdict.

The Civil Justice Is System Is About Resolving Private Disputes, and Only Incidentally About Setting Precedent.

Interestingly, had the jury ruled against Google, the analysis would be just the same, which is to say, the verdict would withstand challenge. That’s right: two different juries listening to the same evidence might have reached different conclusions, and that’s OK. So long as the minds that are disagreeing are “reasonable,” everything’s OK.²²The jury verdict could be challenged in a couple of additional ways. First, the jury instructions could be challenged. The instructions would have be more than just wrong. Their mistakes must have plausibly influenced the jury’s verdict, a tougher standard than you might realize. And the mistaken instruction had to be something Oracle objected to. Second, the court could have wrongly excluded evidence. Here again the bar is high: the evidence must have really have had the potential to change the result. The remedy for either of these bases is a new trial.

That’s a little hard to accept. Either what Google did was OK or it wasn’t, right? Either it’s always OK in all instances to appropriate the organization of APIs, or it’s not, right? But that’s not how it is—just how it appears to interested observers. This is how it really is: there’s a dispute, and the law creates a mechanism for resolving the dispute that we expect/hope is just. And if there’s an appeal, the law might change little bit (or, rarely, a lot).

If there’s something unfair about all this, it’s that the ability to present the best case to a jury requires more money than most people and businesses have. Don’t cry for Oracle: it had plenty of money, and it hired great lawyers, and it had every opportunity to put on a great case. What you need to worry about is when Oracle comes after someone much smaller with hammer and tongs because it can, or when Google infringes the rights of someone much smaller because it can. A theme of the trial was Google’s arrogance, that it felt the old rules didn’t apply to it. The trial showed that, actually, where the parties are evenly matched, the old rules do apply. Where the rules don’t apply is where there’s a disparity between the parties—and the rules don’t apply to the bigger party. $10,000 or even $100,000 in attorney’s fees is a rounding error for some companies; it might also be someone’s life savings.

Next time: are there any lessons to be learned? And, yes, maybe good lawyering matters.

Thanks for reading!

Rick Sanders

Rick is an intellectual-property litigator. He handles lawsuits, arbitrations, emergency injunctions and temporary restraining orders, opposition and cancellation proceedings, uniform dispute resolution proceedings (UDRPs), pre-litigation counseling, litigation avoidance, and other disputes, relating to copyrights, trademarks, trade secrets, domain names, technology and intellectual-property licenses, and various privacy rights. He has taught Copyright Law at Vanderbilt University Law School. He co-founded Aaron | Sanders with Tara Aaron-Stelluto in 2011.