Google v. Oracle - The Final Shoe Drops

The Supreme Court ruled yesterday in Google v. Oracle that Google did not infringe Oracle's copyright in its APIs by virtue of fair use. The vote was 6-2, with Justice Breyer writing for the Court, and Justices Thomas and Alito dissenting. 

The opinion was straightforward and went to great lengths to attempt to explain the technology at issue. I thought it did a decent job of it (definitely more Godot than Guffman), even as the opinion continued to struggle for a good analogy. The Court adopted the file cabinet/drawer/folder analogy presented in Google's brief, which I thought was a terrible analogy...so I guess there's no accounting for taste (or winning advocacy). The court's fair use analysis was influenced by Judge Boudin's concurrence in Lotus v. Borland, though that concurrence didn't actually call it fair use, but instead "privileged use."

Others have and will surely write about the fair use aspects and what this means for software APIs. Contrary to Oracle's ridiculous and vitriolic press statement yesterday, this case will likely not change the way anyone in the industry behaves in the least. APIs have been used and reused for decades, and will continue to be. And contrary to being a barrier to entry, reuse of APIs allows for competitive inroads and entry, including by Oracle, in its mimicry of Amazon's AWS API. (Indeed, the hubris of Oracle's statement in light of its implementation of another company's API is stunning, assuming it was unlicensed-I've been unable to verify one way or the other.)  

The opinion also has some nuggets for other fair use - discussion of transformation and art, definition of markets for determining harm, another reaffirmation of Campbell v. Acuff-Rose Music, fair use as a mixed question of law and fact (something I discussed in a prior blog post), and so forth.

Instead, I will focus on my hobby horse-whether the APIs are copyrighted, and if so how we get to non-infringement. The Supreme Court explicitly decided that the copyrightability of APIs is a third-rail and did not attempt to touch the issue. There are two ways to read the tea leaves. First, perhaps a majority of the court thought they were uncopyrightable, but feared the effects of saying so. Second (and my guess), perhaps a majority of the court (or a 4-4 split) thought that they were copyrightable, but fair use was an acceptable compromise. The second possibility is why I wrote and submitted my amicus brief, which was intended to give a path to non-infringement even if the APIs were copyrightable.

Alas, the court did not buy into the abstraction/filtration argument I made, which I believe was doctrinally appropriate, nor did the brief get a cite, as many that discussed the importance of APIs did. However, in a sense, the court adopted the methodology I suggested. From my brief:

But the copyrightability of an entire work does not answer the question of whether any particular portion of it, if used by another, is infringing. That analysis requires determining whether the defendant has taken too much expression and not ideas, systems, methods of operation, or the like. And such a determination cannot be made outside of the infringement analysis. Any functionally required aspects—including any expression necessary to practice the idea—should be removed from the comparison. 

What remains should then be compared. The advantage of this approach is that it recognizes that while entire software programs may be copyrightable in some contexts, their pieces might not be infringed in others. There need be no zero-sum game, but only a recognition that the scope of copyright depends, as it always has, on the accused’s use of the copyrighted work. 

...

[T]he Court need not decide whether any part of Oracle’s code is copyrightable standing alone. It should only determine that the scope of its copyright in the Java source code cannot extend to infringement through the reuse of declaring functions necessary to create a compiler or interpreter that accepts the same commands and parameter names to allow programmers to use the Java programming language.

The primary pushback on this argument that I received from smart colleagues asked this question: why should context matter in the infringement analysis? Justice Thomas's dissent is replete with this same concern. My answer was always the same - because use in a functional context may be a use of the idea/method, whereas use in a different context might not be.

But Justice Breyer has sidestepped this question to essentially reach the same result using fair use. The one place where we can be sure that context matters is fair use. The nature and character of the use is one of the factors, after all. The Court's analysis tracks many of the same issues in my brief - the functionality of the APIs, their use as a de facto standard, switching costs, etc.

And so the Court's final resolution is not that far off from what I had asked. Rather than excluding the APIs from infringement by filtering them out, the Court would instead exclude them from infringement under a fair use analysis that considers many of the same factors. I can live with this solution--way back in 1999, I published a paper that argued that "courts have been able to determine efficient economic outcomes based on the cases before them, but they have been unable to settle on a rule that definitely determines how much reuse to allow in each case." The article lays out a variety of economic factors that predict how cases come out, and you'll be shocked to find out that they favor Google in this case (e.g. switching costs, de facto standards, lack of slavish copying of the implementation, no breach of an underlying economic duty, public benefits of compatibility). Perhaps that's a reason this case has stuck in my craw for so long: it's about the only one that didn't fit with my 20+ year old model.

Despite my doctrinal sanguinity, the downside of the court's approach is that it might still lead to framing issues in the future. Litigants might still be subjected to juries asked to simply decide whether the APIs were used (for infringement) and then hope to rely on fair use as a defense. Then again, a judicial fair use inquiry might keep the question from ever getting to a jury, which is basically the same result I've advocated. But this case went to a jury, by appellate order, and it's unclear that it should have. It was certainly costly. However, the strong language of this opinion may apply the next time, as in the case I recently blogged about with control codes.

The end is near for my Oracle and Google blogging, a 9 year expedition. But I do have one more in me, a more technical post in a week or so about the so-called 170 lines of code that supposedly all that are necessary to implement Java.

The Rise and Rise of Transformative Use

I'm a big fan of transformative use analysis in fair use law, except when I'm not. I think that it is a helpful guide for determining if the type of use is one that we'd like to allow. But I also think that it can be overused - especially when it is applied to a different message but little else.

The big question is whether transformative use is used too much...or not enough. Clark Asay (BYU) has done the research on this so you don't have to. In his forthcoming article in Boston College Law Review called, Is Transformative Use Eating the World?, Asay collects and analyzes 400+ fair use decisions since 1991. The draft is on SSRN, and the abstract is here:

Fair use is copyright law’s most important defense to claims of copyright infringement. This defense allows courts to relax copyright law’s application when courts believe doing so will promote creativity more than harm it. As the Supreme Court has said, without the fair use defense, copyright law would often “stifle the very creativity [it] is designed to foster.”

In today’s world, whether use of a copyrighted work is “transformative” has become a central question within the fair use test. The U.S. Supreme Court first endorsed the transformative use term in its 1994 Campbell decision. Since then, lower courts have increasingly made use of the transformative use doctrine in fair use case law. In fact, in response to the transformative use doctrine’s seeming hegemony, commentators and some courts have recently called for a scaling back of the transformative use concept. So far, the Supreme Court has yet to respond. But growing divergences in transformative use approaches may eventually attract its attention.

But what is the actual state of the transformative use doctrine? Some previous scholars have empirically examined the fair use defense, including the transformative use doctrine’s role in fair use case law. But none has focused specifically on empirically assessing the transformative use doctrine in as much depth as is warranted. This Article does so by collecting a number of data from all district and appellate court fair use opinions between 1991, when the transformative use term first made its appearance in the case law, and 2017. These data include how frequently courts apply the doctrine, how often they deem a use transformative, and win rates for transformative users. The data also cover which types of uses courts are most likely to find transformative, what sources courts rely on in defining and applying the doctrine, and how frequently the transformative use doctrine bleeds into and influences other parts of the fair use test. Overall, the data suggest that the transformative use doctrine is, in fact, eating the world of fair use.

The Article concludes by analyzing some possible implications of the findings, including the controversial argument that, going forward, courts should rely even more on the transformative use doctrine in their fair use opinions, not less.

In the last six years of the study, some 90% of the fair use opinions consider transformative use.*  This doesn't mean that the the reuser won every time - quite often, courts found the use to not be transformative. Indeed, while the transformativeness finding is not 100% dispositive, it is highly predictive. This supports Asay's finding that transformativeness does indeed seem to be taking over fair use.

Jiarui Liu on the Dominance and Ambiguity of Transformative Use

The Stanford Technology Law Review just published an interesting new copyright article, An Empirical Study of Transformative Use in Copyright Law by Prof. Jiarui Liu. Here is the abstract:

This article presents an empirical study based on all reported transformative use decisions in U.S. copyright history through January 1, 2017. Since Judge Leval coined the doctrine of transformative use in 1990, it has been gradually approaching total dominance in fair use jurisprudence, involved in 90% of all fair use decisions in recent years. More importantly, of all the dispositive decisions that upheld transformative use, 94% eventually led to a finding of fair use. The controlling effect is nowhere more evident than in the context of the four-factor test: A finding of transformative use overrides findings of commercial purpose and bad faith under factor one, renders irrelevant the issue of whether the original work is unpublished or creative under factor two, stretches the extent of copying permitted under factor three towards 100% verbatim reproduction, and precludes the evidence on damage to the primary or derivative market under factor four even though there exists a well-functioning market for the use.

Although transformative use has harmonized fair use rhetoric, it falls short of streamlining fair use practice or increasing its predictability. Courts diverge widely on the meaning of transformative use. They have upheld the doctrine in favor of defendants upon a finding of physical transformation, purposive transformation, or neither. Transformative use is also prone to the problem of the slippery slope: courts start conservatively on uncontroversial cases and then extend the doctrine bit by bit to fact patterns increasingly remote from the original context.

This article, albeit being descriptive in nature, does have a normative connotation. Courts welcome transformative use not despite, but due to, its ambiguity, which is a flexible way to implement their intuitive judgments yet maintain the impression of stare decisis. However, the rhetorical harmony conceals the differences between a wide variety of policy concerns in dissimilar cases, invites casual references to precedents from factually unrelated contexts, and substitutes a mechanical exercise of physical or purposive transformation for an in-depth policy analysis that may provide clearer guidance for future cases.

This article builds on and extends prior empirical work in this area, such as Barton Beebe's study of fair use decisions from 1978 to 2005. And it provides a nice mix of interesting new empirical results and normative analysis that illustrates why fair use doctrine is (at least for me) quite challenging to teach. For example, Figure 1 illustrates how transformative use has cannibalized fair use doctrine since the 1994 Campbell v. Acuff-Rose decision endorsed its use:

Liu also examines data such as the win rate for transformative use over time, by circuit, and by subject matter. But I particularly like that Liu is not just counting cases, but also arguing that courts are using this doctrine as a substitute for in-depth policy analysis.

No Fair Use for Mu(sic)

It's an open secret that musicians will sometimes borrow portions of music or lyrics from prior works. But how much borrowing is too much? One would think that this is the province of fair use, but it turns out not to be the case - at least not in those cases that reach a decision.  Edward Lee (Chicago-Kent) has gathered up the music infringement cases and shown that fair use (other than parody) is almost never a defense - not just that defendants lose, but that they don't even raise it most of the time. His article Fair Use Avoidance in Music Cases is forthcoming in the Boston College Law Review, and a draft is available on SSRN. Here's the abstract:

This Article provides the first empirical study of fair use in cases involving musical works. The major finding of the study is surprising: despite the relatively high number of music cases decided under the 1976 Copyright Act, no decisions have recognized non-parody fair use of a musical work to create another musical work, except for a 2017 decision involving the copying of a narration that itself contained no music (and therefore might not even constitute a musical work). Thus far, no decision has held that copying musical notes or elements is fair use. Moreover, very few music cases have even considered fair use. This Article attempts to explain this fair use avoidance and to evaluate its costs and benefits. Whether the lack of a clear precedent recognizing music fair use has harmed the creation of music is inconclusive. A potential problem of “copyright clutter” may arise, however, from the buildup of copyrights to older, unutilized, and underutilized musical works. This copyright clutter may subject short combinations of notes contained in older songs to copyright assertions, particularly after the U.S. Supreme Court’s rejection of laches as a defense to copyright infringement. Such a prospect of copyright clutter makes the need for a clear fair use precedent for musical works more pressing.

The results here are pretty interesting, as I discuss below.

Thoughts on Google's Fair Use Win in Oracle v. Google

It seems like I write a blog post about Oracle v. Google every two years. My last one was on May 9, 2014, so the time seems right (and a fair use jury verdict indicates now or never). It turns out that I really like what I said last time, so I'm going to reprint what I wrote at Madisonian.net a couple years ago at the bottom. Nothing has changed about my my views of the law and of what the Federal Circuit ruled.

So, this was a big win for Google, especially given the damages Oracle was seeking. But it was a costly win. It was expensive to have a trial, and it was particularly expensive to have this trial. But it is also costly because it leaves so little answered: what happens the next time someone wants to do what Google did? I don't know. Quite frankly, I don't know how often people make compatible programs already, how many were holding back, or how many will be deterred.

Google did this a long time ago thinking it was legal. How many others have done similar work that haven't been sued? Given how long has it been since Lotus v. Borland quieted things, has the status quo changed at all? My thoughts after the jump.