Mass Law Blog

Final Thoughts On Google v. Oracle

Final Thoughts On Google v. Oracle

At long last – after more than ten years, two trials and three appeals – the copyright lawsuit in Google v. Oracle has come to a close. In a surprise ending (given the emphasis on copyrightability for much of the case) on April 5, 2021 the Supreme Court held that Google’s copying of 11,500 lines of code from Oracle’s Java SE Application Programming Interface (the “Java API”) and its use in Google’s Android mobile operating system was copyright fair use.

A deeply disappointing ending for Oracle, which was hoping for a third trial, where it intended to seek ten billion dollars in damages.

I’ve written about Oracle v. Google more than any other case over the last ten years – I count 12 posts (enter “java” in the search bar above to find them or click here and scroll to bottom), and I have a few observations. But first, here is my highly compressed summary of the court’s fair use decision:

(1) The Oracle Java API is a functional “user interface,” analogous to a gas pedal in a car or the QWERTY keyboard. It is entitled to only “thin” copyright protection. (2) Google used the Java API to develop the Android platform software for a smartphone, a “transformative” use. (3) The 11,500 lines of code copied by Google represent less than one-half of one percent of the Java platform. (4) Since many programmers are familiar with the Java API, Google’s copying benefited the public by allowing programmers to use their knowledge and experience to program Android, rather than having to learn a new API. (5) Android did not harm Oracle’s actual or potential markets for the Java API (or so the jury could have found).

What Does Oracle v. Google Mean for the Copyrightability of APIs? For most of the ten years this case was in the courts the central issue was whether Oracle’s Java API was protected by copyright. However, the Supreme Court bypassed that difficult issue – it decided the case on fair use alone. The Federal Circuit’s holding that the Java API is copyrightable remains as precedent in the Federal Circuit, and the Federal Circuit decision can be cited for this holding, with the caveat “reversed on other grounds.” It’s not, however, binding on any other circuit, not even the Ninth Circuit, the circuit from which the case was appealed.

This means that anyone using an API and facing a claim of infringement could still have to relitigate copyrightability and enter the murky waters of fair use – murky because the litigation outcome of fair use is notoriously unpredictable, with a high reversal rate on appeal in the federal circuit courts.

Many companies in the software industry supporting Google had hoped for a ruling that APIs are not protectable, reversing the Federal Circuit. Instead of black and white they have continued uncertainty. This means that it remains risky to use an API without consent, since the user may still be subject to a copyright challenge.

What Does Oracle v. Google Mean for Fair Use When it Comes to APIs? To read the computer press you might mistakenly conclude that copying of APIs is fair use. While Google certainly made it easier to establish fair use in this context, the case doesn’t hold that APIs are subject to fair use in every instance, as a matter of law. For example, not every API will share the popularity of the Java API, and not every case will reflect the same economic non-impact  that was present here. In other words, in a different context Google could be distinguishable. 

What Does Oracle v. Google Mean for Jury Trials in Fair Use Cases? The second trial in this case was a jury trial on Google’s fair use defense. Oracle argued that fair use should be determined by the judge as a question of law de novo, while Google argued that the jury’s verdict should control, and be reversed only if it lacked substantial evidence to justify it. 

The Supreme Court held that fair use is an equitable defense, and therefore should be decided by the judge. The Court didn’t rule out fact finding by a jury, but the judge delivers the last word.

As a practical matter, this means the end of jury trials in copyright fair use cases in the U.S. It’s not inconceivable that a litigant would ask for a jury on fair use, but this would now require a detailed special verdict, and it would make little sense for a plaintiff or defendant to bifurcate responsibility for the case in that manner. It’s far easier to let the judge determine the facts, since the judge will be weighing the four copyright fair use factors. So, juries are likely a thing of the past in fair use litigation. 

Copyright fair use cases have often been decided on summary judgment, and given the responsibility now added to the role of the judge, summary judgment resolutions will be even more common in the future.

What Does Oracle v. Google Mean for Fair Use Beyond Software? Many copyright observers will be tempted to provide opinions on what this case means for copyright fair use, and in particular fair use outside computer software. However, it’s too early for this. This was the Supreme Court’s first major opinion on the fair use doctrine in over 25 years and it will be parsed and applied by the federal district and appellate courts for years to come in ways that are difficult to predict today. Google may broaden fair use law generally, or it may end up being a fact-bound case about functional computer code with little long-term impact for copyright fair use – it’s too early to say.

Nevertheless, it didn’t take long for lawyers to argue the implications of the case beyond software. In the visual art case Warhol v. Goldsmith, decided in March 2021 (just a few days before Google), the Second Circuit held that Andy Warhol’s use of Lynn Goldsmith’s photograph of Prince to create unauthorized silkscreen and pencil artworks was not fair use. Google was decided soon afterwards, and the Warhol Foundation filed a petition for panel rehearing and rehearing en banc, arguing that Google broadens the law of transformative use and public benefit, and established a new balancing analysis for copyright fair use which favors Warhol. As of this post, the Second Circuit has not acted on this petition. 

The Supreme Court Decision Owes a Large Debt to the First Circuit’s Decision in Lotus v. Borland. I can’t read Justice Breyer’s majority decision without seeing it’s debt to the First Circuit’s 1995 decision in Lotus v. Borland, and in particular Judge Boudin’s concurring opinion in that case. To my eyes, in some respects Oracle is an endorsement of Lotus’s “method of operation” holding in the guise of fair use.  

I suspect that Judge Breyer argued that the Court should hold that the Java API was a “method of operation,” and therefore uncopyrightable, as suggested by Lotus, which involved a menu-command user interface. He couldn’t persuade enough justices to back this theory, but when he wrote the opinion he had Lotus in mind, and the decision reflects it.

Google LLC v. Oracle America, Inc. (USSC April 5, 2021)

Final Thoughts On Google v. Oracle

Will the Supreme Court Decide Oracle v. Google on a Technicality?

Will the Supreme Court dodge the thorny copyright infringement issues in the long-running (ten year) Oracle v. Google case on a technicality? The case was originally scheduled to be argued in March 2020, but after Covid-19 it was deferred to the 2020-21 term. Then, on May 4, 2020 the Court ordered the parties to file supplemental briefs:

“The parties are directed to file supplemental letter briefs addressing the appropriate standard of review for the second question presented, including but not limited to the implications of the Seventh Amendment, if any, on that standard. The briefs, not to exceed 10 pages, are to be filed simultaneously with the Clerk and served upon opposing counsel on or before 2 p.m., Friday, August 7, 2020.”

The “second question presented” is Google’s appeal of the Federal Circuit’s 2018 decision reversing a trial jury’s fair use finding in favor of Google. The “standard of review” is a reference to the “de novo” standard used by the Federal Circuit in the opinion under review. The Seventh Amendment is the constitutional right to a jury trial (“the right of trial by jury shall be preserved, and no fact tried by a jury, shall be otherwise reexamined in any court of the United States …”).

The Court gave the parties a long time to write those ten pages – three months. (Oracle’s Brief, Google’s Brief). Now that August 7, 2020 has come and gone, and the letter briefs have been filed, where does the issue stand?.

For brief background let me describe the events that led to this point.

After the Federal Circuit reversed the trial court on the copyrightability of the Java API declaring code there was a jury trial on Google’s fair use defense. The jury entered a general verdict in favor of Google. Oracle filed a motion for judgment as a matter of law before the case went to the jury, and renewed it following the verdict. The trial judge denied that motion, upholding the verdict.

Oracle appealed and in 2018 the Federal Circuit reversed, holding that Google’s fair use defense had failed as a matter of law. The Federal Circuit reviewed the verdict “de novo” – weighing the facts itself and giving no deference to the legal conclusion the jury had reached based its evaluation of those facts. 

That decision (along with the Federal Circuit’s 2014 decision on copyrightability) is now before the Supreme Court.

As it turns out, the issue Oracle and Google focus on in their August 7th briefs is a legal technicality. The technicality arises from the Supreme Court’s conclusion, in a 1985 copyright case, that fair use is a “mixed” question of fact and law. However, mixed questions are not all alike, and may depend on whether a case involves primarily legal or factual work. Which predominates in this case – law or facts? Is Google right that factual issues predominate, in which case the Federal Circuit applied the incorrect standard of review? Is Oracle right that de novo review was appropriate here and that, more broadly, every fair case should be decided by judges alone?


Google argues that the Federal Circuit failed to give appropriate deference to the jury verdict and the trial judge’s post-trial decision in its favor. Google asserts that the correct legal standard, which the trial judge held favored upholding the verdict, is whether a “rational trier of fact” could have reached the jury’s conclusion. Under this standard, Google argues, the jury verdict for Google should be upheld. 

On the mixed fact/law issue, Google argues that in this case factual questions predominated, and therefore the appeals court should have deferred to the jury and trial judge. To drive its point home Google provides this list of factual questions that were presented to the jury: 

    • “the significance of the common practice of reusing software interfaces;
    • the extent to which Oracle made the declarations available to use without a license;
    • whether or how Java SE was used in smartphones or was suited for that environment;
    • how functional the declarations are, and thus removed from the core of copyright;
    • how quantitatively or qualitatively significant the reused declarations were in comparison to the whole of the copyrighted work;
    • how much Sun Microsystems (the original creator of Java SE) supported Google’s reuse;
    • the degree of market harm, if any, suffered by Oracle;
    • how transformative Google’s reuse of the declarations in a smartphone was;
    • whether Google reused more than necessary to achieve an innovative purpose;
    • whether Android competed with Java SE in the market for any derivative product;
    • what were the reasonably likely future derivative markets for Java SE;
    • whether the amount of creativity Android unleashed justified the reuse; and
    • which of the four statutory factors or other unenumerated factors was more or less important in view of all other evidence in the record.”

Google asks the Court to find that when a jury renders a general verdict on disputed facts such as these a reviewing court must view the evidence in the light most favorable to the verdict, not perform its own de novo weighing of the evidence.


Oracle argues that fair use is a primarily legal, not factual, question properly determined de novo:

The ultimate determination whether a defendant’s copying qualifies as fair use entails primarily legal work because that analysis involves legal judgments balancing the competing policies embodied in the Copyright Act of rewarding innovation, protecting the author’s property rights, encouraging progress of science and arts, and safeguarding constitutional free expression. It is the job of a judge, not a lay jury, to calibrate these interests … Because fair use entails primarily legal work, it receives de novo review.

Based on this argument Oracle asks the Supreme Court to go one step further and hold that there is no right to a jury trial on copyright fair use, an issue the Court has never before directly addressed.

And, in case the Court doesn’t accept this argument, Oracle argues that in this particular case the balance of facts and law is mostly legal, and therefore the appeals court was right to review the jury verdict de novo.


First, it’s unlikely that the Court will decide the broad issue of whether fair use should be tried by judge or jury. There’s no need to address such an important issue in this case, particularly when that issue hasn’t been briefed in depth and neither party raised it as a question on appeal. 

However, there is a good chance the Court will decide that under the facts of this specific case the fair use determination was primarily factual, and therefore the Federal Circuit was wrong to review the case de novo – the jury verdict should have been reviewed under the “deferential” standard.

One incentive for the Court to decide the case in this manner is that it allows the Court to avoid the thorny Java API “copyrightability” issue, which is difficult both technically and legally. Deciding the case on fair use grounds avoids that issue. Deciding that the Federal Circuit failed to give proper deference to the fair use jury verdict for Google resolves the case on the narrowest possible grounds.

I’ll go out on a limb and predict that the Court will reverse the Federal Circuit based on its erroneous application of the de novo standard. I’ll also predict that rather than send the case back to the Federal Circuit for a third hearing it will uphold the verdict in favor of Google and end this marathon litigation once and for all. 

Oral argument is schedule for October 7th, so stay tuned. In the meantime, the key legal documents in this case can be found on my Oracle v. Google Resource Page here.

Is the Internet Archive’s “National Emergency Library” Copyright Infringement or Fair Use?

Is the Internet Archive’s “National Emergency Library” Copyright Infringement or Fair Use?

After the Internet Archive launched a “National Emergency Library” the copyright community held its collective breath, waiting to see if the authors and publishers affected would tolerate it, or challenge it in court. Now we have the answer. On June 1, 2020, four major publishers — Hachette, HarperCollins, Wiley, and Penguin Random House — filed a copyright infringement suit against the Archive.

Background. In late March 2020, in response to the COVID 19 pandemic, the Internet Archive opened a digital “library” of 1.4 million books, to last until June 30, 2020 or the end of the emergency in the U.S., “whichever is later.” Anyone, anywhere in the world, can access this online collection. Users can “check out” (download) books for two weeks at no cost, with no limit on the number of copies that can be checked out at any one time. One thousand or ten thousand copies of The Catcher In The Rye could be downloaded and read simultaneously by different users. Authors and publishers receive no payments.

It appears that there was no effort to distinguish books that might be used in an educational setting from those that are unlikely to be used for education. The Internet Archive did not discriminate between books under copyright vs. books in the public domain, or popular books vs. obscure books.

The Archive explained its reason for taking this step as follows:

“to address our unprecedented global and immediate need for access to reading and research materials” . . . [to ensure] that students will have access to assigned readings and library materials that the Internet Archive has digitized for the remainder of the US academic calendar, and that people who cannot physically access their local libraries because of closure or self-quarantine can continue to read and thrive during this time of crisis, keeping themselves and others safe. . . . ‘In a global pandemic, robust digital lending options are key to a library’s ability to care for staff and the community, by allowing all of us to work remotely and maintain the recommended social distancing.’”

The Archive did offer authors and publishers an opt-out/takedown option.

Controversy. The Emergency Library immediately triggered controversy. The Archive asserted that the Emergency Library was protected by fair use. Authors and publishers disagreed. The U.S. Copyright Office analyzed the issue at the request of a U.S. senator.1

The case brought by the publishers was filed in the federal district court for the Southern District of New York – a court, and an appellate circuit, that has deep experience in copyright law, and specifically copyright fair use when applied to digital copies of books.

1.4 Million Books or 127 Books? Despite headlines and press releases, the subject of the publishers’ suit is not the 1.4 million books in the Emergency Library or the Library as a whole. The four plaintiff-publishers do not own theSorry, Library Closed copyright in many of these books. Many of the books may be in the public domain or may be “orphan works” whose owner is unknown. Some may not be registered with the Copyright Office, which is necessary to file an infringement action.

Rather, the complaint alleges infringement of 127 specific books, although this list may be increased by an amended complaint. Absent a class action,2 whether the Archive has infringed these 127 books must be considered on a work-by-work basis. Damages, if a judge or jury decides they are justified, must be decided as to each book individually, not the 127 books as a whole or based on the other 1.4 million books in the collection.

Fair Use. That said, for each book the question of infringement is likely to come down to the Internet Archive’s public justification – fair use. The answer will depend on a court’s (and possibly a jury’s) evaluation of four nonexclusive factors listed in Section 107 of the Copyright Act. The law requires an individualized analysis and weighing of the factors, making the outcome of fair use cases difficult to predict. However, the publishers have chosen carefully – the 127 books include works by authors such as Bill Bryson, Elizabeth Gilbert, Malcolm Gladwell, Erik Larson, Dennis Lehane, C.S. Lewis, Sylvia Plath, J.D. Salinger and Herman Wouk. Most of the books appear to be popular and in-print. Also, these are creative works, which are entitled to strong copyright protection.

Based on these facts, here’s my short take on the factors as applied generally to this group of books.

First Factor: the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes. Under Supreme Court precedent the central issue under this factor is whether the use is “transformative.” A number of cases point to the conclusion that merely reproducing the work in a new format (in this case moving the text from a physical book to digital format) is not transformative.3 The Archives’ lack of commercial motivation may help it on this factor, although the publishers’ complaint argues that in fact there is an underlying commercial motivation.

Because the transformative nature of the use is so important, this factor DISFAVORS FAIR USE.[efn_noteThe Second Circuit’s opinion in the “Google Books” case, Authors Guild v. Google, Inc. (2d Cir. 2015), is unlikely to help the Archive here. In that important case Google created a search engine that provided “snippets” from books, not full text. The court found making the books searchable via a full text index, but limiting access to “snippet views,” qualified as a transformative purpose. The Emergency Library, in contrast, provides unrestricted full text access.[/efn_note]

Second Factor: the nature of the copyrighted work. As noted above, many of the 127 books identified in the complaint are novels and similarly highly creative works. As to these books, this factor DISFAVORS FAIR USE.

Third Factor: The amount and substantiality of the portion used in relation to the copyrighted work as a whole. The books are copied in their entirety. This factor DISFAVORS FAIR USE.

Fourth Factor: The effect of the use upon the potential market for or value of the copyrighted work. It’s not clear that the Emergency Library has hurt sales of the books at issue. I expect this to be a subject of discovery and dispute on a book-by-book basis. Therefore, at present this factor is NEUTRAL.

On balance, under this analysis the Emergency Library is not protected by fair use.

As mentioned, these factors are not exclusive. In theory, a judge could rule, or a jury could find, that the emergency circumstances of a pandemic created a new fair use justification for online book sharing as to any one, or all, of these books. However, there is no legal precedent for such a ruling.4

Controlled Digital Lending. It’s also notable that the suit includes a legal challenge to “controlled digital lending” (CDL) a process that the Internet Archive has permitted for some time and which publishers have objected to but not, until now, challenged in court. Under CDL the Archive lends digital copies of books as long as it owns a physical copy of the book for each copy loaned at any one time. For example, under this “owned-to-loaned” ratio if the Archive has ten copies of Catcher In The Rye in its warehouse, it will loan up to ten copies of the digitized book at any one time. The theory behind this is that it’s analogous to, or at least no more harmful than, the lending practices of a traditional library. The publishers’ challenge to controlled digital lending will require a separate copyright analysis. It appears to be a harder case for the publishers than their challenge to the Emergency Library.7 It may be that the publishers’ goal is to force the Archive to end the Emergency Library. Whether they will also force the Archive to stop controlled digital lending or back off on this issue is to be seen. Controlled lending may have been included as a negotiating card (“kill the Emergency Library but keep the controlled lending for now, without prejudice to our right to challenge it in the future”).

Where’s It All Going? It seems unlikely that the Internet Archive has the financial resources to defend a case of this magnitude, both in terms of defense costs and potential liability.5 Worst case, for the 127 books named in the complaint alone (which could be supplemented), statutory damages could exceed $19 million (127 books x $150,000 per book). The Archive could also be liable for the publishers’ legal fees.

Bottom line: one way or the other, I expect this case to settle quickly.

Update: On June 10, 2020, less than a week after I posted this article the Internet Archive announced that it was terminating the Emergency Library. Controlled digital lending will continue. (link)


A Renter Uses Your House to Film Porno Movies – Copyright Infringement?

A Renter Uses Your House to Film Porno Movies – Copyright Infringement?

I can’t let a decision on this case pass by, both because the facts are so bizarre and because the case is in my backyard, the Federal District Court for the District of Massachusetts.

The plaintiff, Leah Bassett, owns a house on Martha’s Vineyard. She entered into a several-month long lease with Joshua Spafford. Spafford allowed the house to be used to film a number of pornographic movies. Ms. Bassett sued everyone involved, and one of her claims is copyright infringement. She claims that the movies include shots of paintings, slipcovers, wall hangings and the like (over 50 works in total), all of which were created by her.  She asserts that their appearance in the movie scenes violate her copyright rights (reproduction, distribution and public display).

This case received a lot of attention when it was filed. See, for example,What if your house was used in a porn shoot? This homeowner says hers was, and she’s suing (Boston Globe, March 2018)(link);Martha’s Vineyard homeowner says rental was used as porn set (New York Post, March 2018)(link). The case was assigned to District Court Judge Patti Saris, a federal judge highly experienced in copyright law. The defendants filed a motion for summary judgment, and Judge Saris issued her opinion on May 11, 2020 (link). Here’s a short summary of her ruling on the copyright claims.

Copyrightability. With a few exceptions, all of the works are protected by copyright.

De Minimis Doctrine. Maybe – the evidence in the record before the court was insufficient. Ms. Bassett has 45 days to file a document describing which of the works appear in which films, for how long, and in what level of detail. With representative screen shots. (No movies, please).

Fair Use. The first, second and third fair use factors favor Bassett. (Transformative use, nature of the work, and amount copied). The fourth factor (effect on potential market or value of the work) favors defendants. Three to one, fair use defense loses on summary judgment.

Damages. Bassett may be able to establish damages based on a“reasonable fair market licensing fee.” Damages discovery has not taken place, so this option remains open.

Bottom line and Further Thoughts: The defendants still have a chance of dismissing the case as to some of the works, depending on Bassett’s ability to overcome their de minimis defense. However, there will have to be a separate finding on each of the 50-plus works. In the meantime, Bassett has 45 days to compile her evidence on this. I expect it will be a lot of work with a spreadsheet and a stopwatch. With respect to damages, Ms. Bassett never sold her works, so she can’t prove lost profits. The defendants’ profits will be be a challenge, since it will be difficult to show that customers purchased the films because of the background art and allocate part of the profits associated with the films to her work. She hadn’t registered any of the works prior to the infringement, so statutory damages are not available. The judge is right – she is probably limited to “value of use” damages – what the film makers would have paid based on an arms-length licensing deal. However, Bassett may be hard-pressed to establish a licensing fee for her works, since she is not a commercial artist and has no sales/licensing history. And what would the producer of a porno film pay for background art in any case?

My prediction – Ms. Bassett will not go to the effort to catalogue which films her works appear in and how long each work appears in each scene. Her potential damages are not worth the effort. She survived summary judgment on several other counts (infliction of emotional distress, Chapter 93A and defamation, in particular). Including a copyright case over 50 works will only make the trial longer and more complicated, and distract the jury from the more sympathetic claims where she has a good shot at substantial damages. My two cents: discretion is the better part of valor, and Ms. Bassett will be better served by exercising discretion and dismissing her copyright claims. Bassett v. Jensen (D. Mass. May 11, 2020)

Update: I was mistaken – the plaintiff did submit additional detail on copyright infringement to the court, and on August 6, 2020 Judge Saris entered summary judgment with respect to ten items. (link, via Evernote) Presumably (and absent settlement) the case will now proceed to trial on copyright damages, along with the other claims asserted by the plaintiff-home owner.

Supreme Court Focuses on Standard of Review/Seventh Amendment Issue in Oracle v. Google

Supreme Court Focuses on Standard of Review/Seventh Amendment Issue in Oracle v. Google

The odds of Oracle coming out on top in the Supreme Court appeal of Oracle v. Google just took a turn for the worse.

On May 4, 2020 the following entry appeared on the Supreme Court docket in the long-pending Oracle v. Google copyright case:

The parties are directed to file supplemental letter briefs addressing the appropriate standard of review for the second question presented, including but not limited to the implications of the Seventh Amendment, if any, on that standard. The briefs, not to exceed 10 pages, are to be filed simultaneously with the Clerk and served upon opposing counsel on or before 2 p.m., Friday, August 7, 2020. (Emphasis added)

The “second question presented” is Google’s appeal of the Federal Circuit’s decision reversing a trial jury’s fair use finding in favor of Google. The “standard of review” is a reference to the “de novo” standard used by the Court of Appeals for the Federal Circuit (CAFC) in the opinion under review. The Seventh Amendment is the constitutional right to a jury trial (“the right of trial by jury shall be preserved, and no fact tried by a jury, shall be otherwise reexamined in any court of the United States …”).

This order appeared without warning after the Supreme Court bumped oral argument in this copyright case to an as yet unspecified date in the 2020 term. The case had been scheduled for argument on March 24, 2020, and the Court removed it from the calendar only 12 days before that date.

We aren’t told what prompted the Court to ask for additional briefs in this case, but the order brought to mind one of the amicus briefs filed in support of Google. The brief (link), filed by several law professors (the “Professors Brief”), focused on precisely the issued on which the Court requested briefing. The opening section of the brief summarized their argument –

Reversal of a jury verdict on the issue of fair use is extraordinarily rare. For two centuries, courts have given great deference to jury verdicts. Indeed, history overwhelmingly demonstrates that juries are uniquely situated to make the discretionary judgments that fair-use cases call for. But in this case, the Federal Circuit ignored history, along with the law. It applied de novo review to overturn the jury verdict of fair use. This is the first time that has ever happened. And it is unconstitutional. The Seventh Amendment requires that a jury finding of fair use not be “re-examined” under a de novo standard of review.

This is a powerful argument. It may be Google’s best argument now that the Court has (if belatedly) focused on this issue.

A Brief Recap. In this case following a trial, appeal and remand on infringement a jury trial was held solely on the issue of fair use. After a two week trial the jury returned a verdict for Google, finding that Google’s copy of the Java application programming interface (the Java API) was protected by fair use. Oracle appealed to the CAFC.

On appeal the CAFC held that while fair use is a mixed question of fact and law, this did not “dictate the applicable standard of review.” The CAFC concluded that in deciding copyright fair use the jury’s role “is limited to determining disputed ‘historical facts, not inferences or conclusions to be drawn from those facts.” Thus, “all jury findings relating to fair use other than its implied findings of historical fact must, under governing Supreme Court and Ninth Circuit case law, be viewed as advisory only.” Accordingly, the CAFC held that it must “assess all inferences to be drawn from the historical facts found by the jury and the ultimate question of fair use de novo” – without reference to any legal conclusion made by the previous court to hear the case.

Based on de novo review the CAFC set aside the jury verdict for Google and decided the copyright fair use issue in favor of Oracle. In other words, despite the jury verdict it decided that Google’s use of the Java API is not protected by fair use.

The CAFC’s decision was surprising, to say the least. Appellate reversal of a jury finding on fair use is virtually unprecedented in the law. A determination fair use at trial is based on a weighing of the four fair use factors, and the CAFC was wrong to abrogate the jury’s conclusion on this issue. Or so the argument went.

However, in the three merits briefs filed with the Supreme Court Google and Oracle gave short shrift to the standard of review, at least relative to the other issues. They discussed the Seventh Amendment not at all.

Clearly the Supreme Court has concluded that this issue deserves more attention.

Does the Standard of Review/Seventh Amendment Argument Have Legs? Until we see the parties briefs in August the Professors Brief may be the best guide on this issue. The Professors ground their argument in 18th and 19th Century English common law, their point being that English law courts heard issues similar to modern fair use, and that under English law the courts left the issue of fair use to the jury to decide. This is not as far-fetched as it might seem at first – the Supreme Court has often looked to English common law as an antecedent to help it interpret and apply U.S. copyright law.

From there they move to the Supreme Court’s 1998 decision in Feltner v. Columbia Pictures Television where, tying modern U.S. copyright law to English common law precedents, the Court held that the Seventh Amendment provides a right to a jury trial on all issues pertinent to an award of statutory damages. Fair use, the Professors argue, is pertinent to an award of statutory damages, since statutory damages are a potential remedy in a case premised on a fair use defense. That being the case, the Seventh Amendment applies to a jury determination on fair use, implying a deferential (not de novo) review of a fair use jury verdict.

Lastly, the Professors argue that Supreme Court precedent requires a deferential standard of review in cases involving a mixed question of law and fact (such as fair use), where the issue entails “primarily . . . factual work.” U.S. Bank National Ass’n v. Village at Lakeridge, LLC. The Professors argue that the  copyright fair use determination in this case rested upon its individualized facts, and therefore primarily involved factual work. Accordingly, the CAFC erred in declining to apply a deferential standard of review to the jury verdict.

Did Oracle Make a Mistake By Agreeing That The Jury Could Issue a “General Verdict”? It’s not uncommon for juries to be asked complete detailed “special verdicts”  that may shed light on the the bases for their decisions. For example, in a copyright fair use case the jury might be asked to indicate how it decided each of the fair use factors in reaching its decision.

But that didn’t happen in this case. Instead, Oracle and Google agreed that the jury need only issue a “general verdict.” The jury made no specific findings of fact.

Whether this was a strategic mistake by Oracle or part of a calculated trial strategy we don’t know, but at this point it appears to put Oracle at a disadvantage. The general verdict could weigh in favor of a finding that the CAFC erred in applying a de novo standard, since there were no “historical facts” for the CAFC to analyze, short of the CAFC reviewing the entire trial record and finding facts without the ability to evaluate witness credibility.

What Next? We will have to see what the parties have to say on this issue when they file their briefs in August. However, a decision based on a narrow procedural ground such as the standard of review is likely to be attractive to the Supreme Court. It allows it to avoid the mystifying complexities of copyright law as applied to computer software technology. It allows the Court to avoid revisiting the law of copyright fair use, a doctrine the Court has not addressed in-depth in the 26 years since it decided Campbell v. Acuff-Rose Music, Inc. It enables it to decide the case on a narrow standard-of-review issue and hold that a jury verdict on fair use should not be reviewed de novo on appeal, at least where the jury has issued a general verdict.

And, potentially (should the Court decide to resolve the case on this ground per curiam without hearing before the start of the 2020 term), it is one less case from the 2019 term that the Court would need to carry over into the 2020 term.

It’s worth noting that such an outcome would be disappointing to many people in the computer industry who are hoping for a reversal of the CAFC’s 2014 holding (on an earlier appeal) that the Java API is protected by copyright. And, while Google would be happy with this outcome, it’s not certain that it would be a clear winner for Google and end this case once-and-for all. It’s likely that the Court would decide that the jury verdict should stand, ending the case. But there’s also a risk that it could send the case back to the CAFC for yet another appeal, to be decided under a more deferential standard of review, but one that (given the CAFC’s hard-to-ignore antipathy to Google in this case), still leaves Google at risk.

It’s also worth noting the unusual sequence of events in this case. But for COVID-19’s disruption to the Court’s schedule, argument on this case would have been heard on March 24, 2020. The Court had not requested supplemental briefing prior to that date, and the case would have been taken under advisement on the thin record (as to this issue) as it existed at that time. The delay gave the Court time to request additional briefing on the standard of review, and the risk of that action falls entirely on Oracle.

Final Thoughts On Google v. Oracle

Oracle v. Google: Will The Best Analogy Win?

Analogies, it is true, decide nothing, but they can make one feel more at home – Sigmund Freud

One good analogy is worth three hours discussion – Dudley Field Malone

Oracle v. Google, now before the Supreme Court, is a complicated case in more ways than one. The copyright law issues are difficult, but the case is made even more challenging by its subject matter, which involves highly technical and abstruse computer technology. Judges have a hard enough time applying copyright law to traditional media like music, novels and photographs, but software copyright cases add another order of magnitude of complexity.

The legal briefs now before the Supreme Court are overflowing with computer jargon. You can read all about the “Java language,” the “Java virtual machine,” the “Dalvik virtual machine,” “application programming interfaces (APIs),” “packages,” “classes,” “calls,” “declarations,” “methods,” “implementing code” and “declaring code,” and much more.

Nor did the Federal Circuit pull any punches in its 2014 decision (one of two under appeal.) The Federal Circuit described the Java system as follows:

In the Java system, source code is first converted into `bytecode,’ an intermediate form, before it is then converted into binary machine code by the Java virtual machine” that has been designed for that device. The Java platform includes the Java development kit (JDK), javac compiler, tools and utilities, runtime programs, class libraries (API packages), and the Java virtual machine.

If you’re shaking your head with confusion (or your eyes are glazing over), you’re probably not a Java programer. These are difficult concepts, especially when they are encountered for the first time by people with no background in the complex world of the Java runtime environment.

As best I can determine, the nine Supreme Court justices have little or no background in computer technology. It’s unlikely many of the Supreme Court justices have seen a computer program, less likely they have written one, and even less likely that they are familiar with the Java programming environment. Even if the younger judges received some exposure to programming in college, the world of Java programming is a far cry from the days of BASIC, Fortran and Cobol, the languages taught when the justices were in college.

If, as commentators have suggested, the Federal Circuit’s 2014 decision reflects “a fundamental misunderstanding of how software works” where does this leave Google and Oracle, who have to explain the Java environment to the justices so that they can understand the Java API and decide whether Google infringed it?

One answer is through analogy and metaphor. The right analogy has the potential to go to the heart of the case and persuade a judge who might otherwise feel uncomfortable with the technicalities of the case. The right analogy can make the difference between winning and losing this case. However, analogies can also be dangerous – an analogy may be distinguished or even turned against the offering party, so an advocate has to use them carefully.

The last (and until now only) computer software copyright case heard by the Supreme Court in 1996 — Lotus v. Borland — demonstrates this. That case, which involved an easy to understand menu command hierarchy, was far simpler than Oracle/Google from a technical standpoint. Nevertheless, analogies played a significant role. The First Circuit compared Lotus’s menu commands to the buttons used to control a VCR. During oral argument before the Supreme Court the analogies included the dashboard of a Model T Ford (Justice Souter), a system for organizing a department store and labels on the controls in a plane’s cockpit (Justice Breyer), and a method of dance notation and a language (Borland’s attorney).

There has been no shortage of attempted analogies as Oracle v. Google has made its way through the courts over the last ten years. However, in its Supreme Court merits brief filed on January 6, 2020, Google seems to have forgotten the power of analogy. Beyond a simple filing cabinet analogy the district court judge used and that neither party disputes (each package is like a filing cabinet, each class like a drawer and each method like a folder), Google’s brief is lacking analogies that go to the heart of what an API is and how it relates to the Java implementing code.

Fortunately for Google, Google’s supporting amici came to the rescue. Or, it would be more accurate to say that they tried – there’s no guarantee that the justices actually read amicus briefs (there are 27 of them so far, and that number will increase after Oracle files its merits brief).

Nevertheless, the analogies posited by the Google amici are potentially powerful. Here is a sampling.

The Juke Box Analogy. Few academics have written more about copyright protection of APIs than Professor Peter Menell. The Brief of Professors Peter S. Menell, David Nimmer and Shyamkrishna Balganesh (link), written by him, is an argument in support of Google wrapped in a treatise explaining the history of software copyright law. Here is my favorite example from this brief:

If the Java programming language is analogized to musical language, each API implementation can be characterized as a record album featuring songs (methods). Java Standard Edition (SE) then functions like an electrical-mechanical juke box, containing API record albums from which programmers can choose particular songs by invoking declarations (song titles). The fact that another juke box uses those song titles (declarations) to invoke a known song (method) is purely functional: it does not copy a song (method), it merely identifies a known song (method).

I suspect that if there is one amicus brief the justices may want to read it is this one, given that the Nimmer treatise caries so much authority on copyright law.

The Car Controls Analogy. The Brief Amici Curiae of Eighty Three Computer Scientists (link). borrows from the First Circuit’s decision in Lotus v. Borland and uses the analogy of the controls of a car:

A steering wheel and gas and brake pedals have been standard in cars for over a century. . . . Treating software interfaces as copyrightable would be like requiring car manufacturers to invent a substitute for the steering wheel. Startups would not risk manufacturing such a car, and even if they did, consumers likely would not purchase it.

This analogy never gained much traction in the many software copyright cases decided post-Lotus, and I doubt that it will be persuasive here.

The New York City Map Analogy. IBM and Red Hat’s amicus brief (link) analogizes the Java software interface to the New York city maps at issue in a 1879 Supreme Court case, Perris v. Hexamer, where the Court held that a copyright in maps did not extend to a “system of coloring and signs” for identifying real property characteristics or to a “key” which explained symbolic meanings of coloring and signs. Interestingly, Google does not cite this case, even in passing.

The Supreme Court Language Analogy. The entire amici brief of the Empirical Legal Researchers (link) is devoted to developing and explaining an analogy. The brief states:

The facts of this case are unusually technical and threaten to obscure the legal issues. As attorneys who are also software developers, the amici offer an analogy between the computer languages, with which the Court may be unfamiliar, and a kind of language with which the Court is uniquely familiar: the text of Supreme Court opinions.

You’ll have to read the brief if you want to understand this complex analogy and the statistic analysis on which it is based. Perhaps it’s accurate, but it didn’t hit home with me.

The Book and Title Analogy. The Software and System Developers and Engineers for U.S. Government Agencies amici brief (link) uses the analogy of the relationship between a book and its title:

For developers, relying on a declaration to identify a component for interoperation is like relying on a book title to refer to a book, enabling someone to identify the book to locate it and read it. While a book, like an entire software program, may be subject to copyright, a title of a book has long been understood as not subject to copyright.

This is the simplest and best analogy that I’ve seen in the amicus briefs. It will be interested to see if any of the Justices raise it at oral argument.

The Online Checkout Analogy. The amicus brief of the Small, Medium and Open Source Technology Organizations (which includes Mozilla, Shopify, Etsy and Wikimedia Foundation; link) encourages the Court to view software interfaces as “similar to electronic checkout forms you see when shopping online,” where the fields and structure, sequence and organization have become standard conventions. I’m not sure I agree with this analogy, but I can’t challenge the technical qualifications of the companies that proposed it, so perhaps it has some merit.

What about Oracle? How will it respond to these analogies?

We won’t know that until Oracle files its merits brief in February. But Oracle hasn’t been shy in using analogies in this case. Oracle’s 2014 opening brief to the Federal Circuit (link) began with this lengthy analogy:

Ann Droid wants to publish a bestseller. So she sits down with an advance copy of Harry Potter and the Order of the Phoenix—the fifth book—and proceeds to transcribe. She verbatim copies all the chapter titles—from Chapter 1 (“Dudley Demented”) to Chapter 38 (“The Second War Begins”). She copies verbatim the topic sentences of each paragraph, starting from the first (highly descriptive) one and continuing, in order, to the last, simple one (“Harry nodded.”). She then paraphrases the rest of each paragraph. She rushes the competing version to press before the original under the title: Ann Droid’s Harry Potter 5.0. The knockoff flies off the shelves.

J.K. Rowling sues for copyright infringement. Ann’s defenses: “But I wrote most of the words from scratch. Besides, this was fair use, because I copied only the portions necessary to tap into the Harry Potter fan base.”

Obviously, the defenses would fail.

Defendant Google Inc. has copied a blockbuster literary work just as surely, and as improperly, as Ann Droid—and has offered the same defenses.

The analogy between the Java API and the topic sentences of a fictional literary work seems far fetched. Nevertheless, Oracle seems very attached to it, so look for Oracle to use it again in its Supreme Court brief.

I have no doubt that the many lawyers and technologists on both sides of this case have devoted hundreds, if not thousands, of hours searching for the perfect analogy – the analogy that will go to the heart of the case, make sense to the justices, and lead to victory for their side. Whether Google or Oracle will reach this goal remains to be seen. But when the transcript of oral argument is released  keep an eye open for analogies used by the justices and the parties to see which were used and whether they appeared to be effective.

By the way, if you’ve read this post and you’re still wondering, “what the heck is an application programming interface?” you may want to start with Sara Jeong’s 2016 article, What an API Is and Why It’s Worth Fighting For” (link).

*    *    *

Postscript: Of the nine justices serving on the Supreme Court when Lotus v. Borland was argued in 1996, only three remain today: Ruth Bader Ginsburg, Clarence Thomas and Stephen Breyer. Justice Breyer has a particular interest in copyright law, and he was the most active questioner during oral argument in that case. Pay close attention to him during oral argument in this case.

For a summary of all of the amicus briefs filed to date see Jonathan Brand’s post: Broad Support for Google in the First Round of Supreme Court Briefing

Update, 2-13-2020: Oracle did use the Harry Potter analog in its merits brief, but a much shorter version than it used at the Federal Circiuit:

By Google’s logic, a plagiarist could define J.K. Rowling’s idea as “a story about Harry Potter, Ron Weasley, and Hermione Granger who attend Hogwarts” and steal the characters and their back stories. Or she could market detailed knock-offs of bestsellers by declaring that she “had no other choice” but to reproduce verbatim the 11,300 most memorable sentences or scenes because they were “necessary” to allow fans to use their existing knowledge

Final Thoughts On Google v. Oracle

Oracle v. Google In a Nutshell

Oracle’s copyright case against Google has dragged on for nine years. The case has generated multiple federal district court trials and appellate decisions. Hundreds of thousands of words have been written on the case. Academic careers have been built on it (OK, I’m exaggerating, but not by much).

Now that the case is before the Supreme Court a new, even larger audience wants to understand it. However, few people want to struggle through the lengthy court decisions or law review articles.

Here is my summary of the issues in the case in a nutshell. Almost all jargon and many details omitted. 


First Issue – copyrightability. Oracle owns the Java programming language. Part of Java is an application programming interface (“the Java API”). These are pre-written programs that allow programmers to perform common programming tasks. When Google built its Android smartphone operating system it copied verbatim a significant portion (over 11,000 lines) of the Java API.

The Java API uses commands and syntax that look a lot like computer code. Here’s an example:

The Copyright Act protects computer programs. It defines a computer program as “a set of statements or instructions to be used directly or indirectly in a computer in order to bring about a certain result.” Oracle argues that both the text and the structure, sequence and organization of the API fall squarely within this definition.

However, the Copyright Act excludes from copyright protection “methods of operation,” a term that is undefined. Google argues that the Java API is a method of operation, and therefore not copyright-protected.

The Federal Circuit held that the text and the structure, sequence and organization of the Java API are protected by copyright. Google is asking the Supreme Court to reverse that holding.

Second issue – fair use. Even if the Java API is copyright-protected Google has a second defense – that it’s use of the Java API is fair use. A jury decided this issue in Google’s favor. Oracle appealed, and the Federal Circuit reversed, holding that Google’s use of the Java API is not protected by fair use. Google is asking the Supreme Court to reverse that holding.

What happens next. If the Supreme Court holds that the Java API is not copyright-protected, Google wins and the case ends. If it holds that Google’s use of the Java API is fair use Google wins and the case ends. 

If the Court rules in favor of Oracle on both issues (upholding the two Federal Circuit decisions) the case will be remanded for a trial in the district court to determine Oracle’s damages. Oracle is expected to ask a jury to award it $9 billion.

That’s it, in a nutshell.

If you want to explore the case in more detail go to my Oracle v. Google Resources Page.

A Few Observations From the Ninth Circuit En Banc Argument in Skidmore v. Led Zeppelin

A Few Observations From the Ninth Circuit En Banc Argument in Skidmore v. Led Zeppelin

I have a few observations on the Ninth Circuit September 23, 2019 en banc hearing in Skidmore v. Led Zeppelin. Video of the oral argument is embedded at the bottom of this post, and the transcriptions below are mine – I’ve left out a few words here and there to make this easier to read, but I didn’t leave out anything material.

Did Skidmore’s Attorney Give Away the Case?

Quite possibly.

Here are the key excerpts from the oral argument. (I’m labeling all of the judges’ questions as simply “judge,” but the questions were posed by different judges):

Judge: Are you conceding today that if you are confined to the deposit copy your copyright claims are not viable?

Skidmore Counsel: I think that it is very difficult for plaintiff to win based on the deposit copy since it’s such an inaccurate transcription of the composition ….

Judge: Is that a “yes”? …

Skidmore Counsel: Yes, I think that is the reality of the situation.

Judge: When you listen to these two recordings, the deposit copy and the performed copy by Led Zeppelin, I don’t see how any juror could find they’re substantially similar, even with expert testimony. So you’ve got to get your sound recording in to win this case, don’t you? . . .

Judge: If the law is that the deposit copy is the four corners of the copyright deposit there’s nothing else to retry, correct?

Skidmore Counsel: If that’s the law ….

Judge: And you lose the case unless [the jury hears the sound recording]; a hundred times out of a hundred, right? You’ve gotta get your sound recording in in order to win this case, don’t you?

Skidmore Counsel: I think so.

Later in the hearing, when Led Zeppelin’s attorney was arguing:

Judge: It seems to me that assuming the copyright analysis is limited to the deposit copy any error [in the jury instructions] is harmless because no reasonable juror could find that the alleged copying was unlawful appropriation of the deposit copy, so why could we assume that instructional error occurred and find any error harmless?

Counsel for Led Zeppelin: I agree wholeheartedly.

The problem here, from Skidmore’s perspective, is that the judges appeared unconvinced that the sound recording of Taurus should be admissible.

This sets up the scenario (referenced by the judge in the quote above) where the en banc court would not need to address the jury instruction issue that was grounds for reversal by the Ninth Circuit panel, and which at least one judge indicated was problematic. The court can bypass that as “harmless error.”

Whether the Ninth Circuit will take this easy way out remains to be seen, but it seems like a significant risk for Skidmore.

By the way, I’m being facetious when I ask whether Skidmore’s attorney “gave away the case.” As a strategic matter he may have had no choice. He may have concluded that he cannot win a retrial on the deposit copy (he lost the first trial when limited to that evidence), and he doesn’t want a remand on a technicality such as deficient jury instructions if he will be limited to that evidence.

The “Inverse Ratio” Rule

The inverse ratio rule provides that if the plaintiff establishes a high degree of access to its work by the defendant, a finding of copyright infringement may be based upon a lesser degree of similarity. This controversial doctrine exists only in the Ninth Circuit. The trial judge refused to give the jury an inverse ratio instruction, an issue Skidmore raised on appeal.

This doctrine may already be dead as a practical matter. In the 2018 Blurred Lines case (Gaye v. Williams) the original opinion suggested that there was some validity to this doctrine, and asked the trial court to reevaluate it on remand.1 But the Ninth Circuit issued an amended opinion deleting all references to the rule, arguably implying that it did not endorse it.

At the en banc hearing, one of the judges asked Led Zeppelin’s lawyer to address this doctrine, and the lawyer briefly explained why it was bad copyright law (a view widely shared in the copyright community). Quite possibly, this case will allow the Ninth Circuit to put this much-criticized doctrine to bed, once and for all.

The “Thin Copyright” Issue

As I discussed in some detail recently (Copyright Office Backs Led Zeppelin In Ninth Circuit En Banc Appeal), the United States filed an amicus brief urging the court to adopt a “virtually identical” test for musical works that contain only “a small number of standard elements” (as is the case with Taurus, they argued).

An attorney from the Department of Justice argued for the United States at the hearing, and frankly he didn’t push this issue as hard as he might have. He certainly didn’t propose a rule that the Ninth Circuit could adopt, and the judges appeared unreceptive to this argument. I’d be surprised if a new copyright rule establishing a thin copyright for some musical works came out of this case.


Courts are notorious for creating the wrong impression during oral argument. Many a lawyer has walked out of a hearing after getting knocked around by a judge thinking that she had lost, only to win when the ruling arrives. And vice versa.

With that caveat, I will venture a prediction that the Ninth Circuit will hold that (a) the copyright in pre-1978 works is limited to the deposit copy, and (b) Skidmore can’t prevail on the deposit copy. Accordingly, any errors in the jury instructions were harmless and the jury verdict of non-infringement is affirmed.


Copyright Office Backs Led Zeppelin In Ninth Circuit En Banc Appeal

Copyright Office Backs Led Zeppelin In Ninth Circuit En Banc Appeal

Update (9/25/19): A Few Observations From the Ninth Circuit En Banc Argument in Skidmore v. Led Zeppelin (link)

The appeal in Skidmore v. Led Zeppelin is scheduled to be reargued before an en banc Ninth Circuit appeals court panel on September 23, 2019 (watch it live online here), and the U.S. Copyright Office has taken the unusual step of submitting an amicus brief in support of Led Zeppelin.1

This important copyright case is discussed in my October 2018 post, Led Zeppelin, Spirit and a Bustle at the Ninth Circuit, so I won’t review the background in detail here. The works at issue are Spirit’s 1968 song Taurus2 and the opening section of Led Zeppelin’s Stairway to Heaven. A Ninth Circuit panel reversed the jury’s verdict (verdict here) in favor of Led Zeppelin and sent the case back for retrial based on errors in the jury instructions. Led Zeppelin petitioned the Ninth Circuit to hear the case en banc, and the court granted this request. This means that the appeal will be reheard by the chief judge and ten other Ninth Circuit judges.3 

The Ninth Circuit grants only about 1% of en banc petitions, and the Copyright Office rarely files an amicus brief in cases that have not reached the Supreme Court.4 The fact that both events have converged in this case underscores its importance. The music and copyright communities are watching closely.

The Copyright Office amicus brief makes two arguments.


Pre-1978 Unpublished Musical Works – Is the Sound Recording Covered By the Copyright? This has been an issue in several recent music copyright cases, and I’ve written about it in connection with the Blurred Lines and Ed Sherran cases, as well as the first Led Zeppelin appeal, decided by a 3-judge Ninth Circuit panel. The panel held that the copyright in unpublished works under the 1909 Act were defined by and limited to the deposit copy, not the sound recording. The Copyright Office urges the en banc court to reach the same conclusion.

This argument is based on an arcane copyright law technicality. However, the Copyright Office’s position is strong, and I expect the en banc court to reach the same conclusion that the panel reached. 

“Substantially Similar” or “Virtually Identical”? Copyright protection for pre-1978 sound recordings may be an important issue for Skidmore and Led Zeppelin, but the vast majority of musical works that are the focus of infringement claims involve current/post-1978 works. Therefore, the second issue addressed by the Copyright Office — whether the legal standard for infringement should be “substantially similar” or “virtually identical” —  is far more important to the music community.

At the trial of this case the presiding judge instructed the jury to determine whether there was copyright infringement based on the “substantial similarity” test. At trial, and in the first appeal of this case, Led Zeppelin’s lawyers argued that the musical elements from Taurus that were at issue should be protected by only a “thin” copyright, and therefore the test for infringement should be “virtually identical copying.” The premise underlying this argument is that Taurus is little more than a “selection and arrangement” of unprotectable musical elements, an argument that Skidmore vehemently disputes. 

The Ninth Circuit panel did not address this argument in the Led Zeppelin appeal. However, it subsequently rejected it in Williams v. Gaye (the Blurred Lines case), stating:

We reject the Thicke Parties’ argument that the Gayes’ copyright enjoys only thin protection. Musical compositions are not confined to a narrow range of expression . . . [A]s we have observed previously, music … is not capable of ready classification into only five or six constituent elements, but is instead comprised of a large array of elements, some combination of which is protectable by copyright . . . We have applied the substantial similarity standard to musical infringement suits before, . . . and see no reason to deviate from that standard now. Therefore, the Gayes’ copyright is not limited to only thin copyright protection, and the Gayes need not prove virtual identity to substantiate their infringement action.

The Copyright Office attempts to distinguish Taurus, urging the Ninth Circuit to give Taurus only thin protection, apply the “virtually identical” test, and affirm the jury verdict:

[Williams v. Gaye] did not involve an allegation that the infringing work copied only an arrangement of a small number of standard elements . . . the court did not consider the scope of protection for a single phrase comprising a small number of basic musical elements. The opinion should not be read as requiring more than thin protection for a portion of a musical work with such limited originality . . ..

Fundamentally, then, the allegedly copied material, as embodied in the deposit copy of Taurus, consists of the use and placement of a combination of basic and not copyrightable elements: the A-minor chord and the descending chromatic scale. While these elements have been selected and arranged to create the bass line pattern at issue, this particular combination, employing simple rhythm and the standard technique of arpeggiation, is itself relatively simple. Such a combination is subject, at most, to a thin copyright; there are not many ways to express this type of chromatically descending arpeggiated bass line and have it still be recognizable as such. Providing broad protection would effectively result in the protection of a musical idea, granting a monopoly on a common musical convention. Infringement should therefore occur only if the allegedly copied portion appears virtually identically in the two works.5

Not surprisingly, Skidmore takes a contrary view. He has filed a supplemental brief, responding to the Copyright Office’s amicus brief, arguing that “‘thin copyright ‘applies not to the ‘broader protection accorded artistic works,’ but instead to functional, fact-driven, or scenes a faire copyrights where there are a limited number of ways to express the ideas contained in the work.” In another filing (responding to the amicus briefs filed by intellectual property professors and musicologists), Skidmore asserts that “nowhere has this Circuit, or any Circuit, ever applied ‘thin’ copyright or virtual identity to music copyright cases, and have in fact at all points specifically warned against applying the concept to such artistic works.”

How the Ninth Circuit will handle this dispute is important to the future of music copyright law. However, the most immediate problem with the Copyright Office’s argument is that it is asks the appeals court to make a finding that the trial judge declined to make. The trial judge’s jury instructions make clear that while the jury could consider Taurus’s non-original or public domain elements in its deliberations, the judge did not conclude that Taurus was limited to those elements. Accordingly, the jury instructions use the “substantial similarity” test. Although its possible, it seems unlikely that the appeals court will make a ruling that the trial judge declined to make following courtroom testimony by experts on both sides of the case.

However, the issue raised by the Copyright Office needs to be viewed in a broader perspective. It is beyond dispute that copyright law struggles with music infringement. The broad application of copyright principles to works of varying types is one of copyright’s strengths, and has given copyright the ability to adapt to new forms of expression, from photography in the late 19th century to software today. However, these principles face a challenge when it comes to music, where copying and borrowing between musicians is an accepted part of musical practice and the venerable copyright law “idea/expression” distinction has little or no meaning.

The Ninth Circuit could resolve the two extremes presented by the Copyright Office and the parties in this case (“substantial similarity” vs. “virtual identity”) with a more flexible legal test that (except in the most extreme cases), leaves the decision up to the jury. Under this test the judge would instruct the jury that to the extent the jury finds the plaintiff’s music to be comprised of basic, common or public domain elements the jury should give it “weaker” protection. At the other extreme protection should be stronger. However, it would not be a judge-determined either/or test (as is the law today), but a flexible test where the jury would apply the standard on a continuum between these two poles.

 For example, Rick Beato makes a convincing case that Taurus uses a musical phrase contained in many pre-Taurus popular and classical pieces of music. If the jury is persuaded by an expert presenting this evidence, it would be free to conclude that the copyright in Taurus is thin, and judge Stairway To Heaven based on the “virtually identical” standard, or toward that end of the continuum. If Skidmore’s expert had persuasive evidence and opinion to the contrary, the jury might apply a standard more in the direction of “substantially similar.” However, in all but the most obvious cases that would be a decision for the jury, not a judge.

Beyond this, it would be welcomed if the Ninth Circuit took this opportunity to clarify or simplify music copyright law, which must be incomprehensible to most jurors. A glance at the jury instructions in Skidmore v. Led Zeppelin illustrates this (see, in particular instruction 21, which asks the jury to apply the dreaded Ninth Circuit “extrinsic/intrinsic similarities” test).6 It’s indisputable that the jury instructions in copyright cases are head-scratchingly complex to jurors (and to many lawyers). It’s not fair to musicians or jurors to require them to apply a body of law that has devolved to this level of complexity.

Sadly, the Ninth Circuit is known for moving in the direction of greater complexity, not less. Regardless of the outcome of this particular case, the odds are that the en banc decision in Skidmore v. Led Zeppelin will leave music copyright law in an even worse state than it is today.


Copyright Infringement? Peloton Punches Back With Antitrust

Copyright Infringement? Peloton Punches Back With Antitrust

[This post was updated in September 2019:  “Contract, Combination or Conspiracy” – Can Peloton’s Lawsuit Survive the Music Publishers’ Motion to Dismiss?]

Can a trade association negotiate sales or licenses on behalf of its members? Can it tell members, “don’t negotiate individually with a specific purchaser, and if you are already in negotiations with that purchaser cut them off and let us negotiate on behalf of you and other members”? At what point does this conduct become an antitrust violation? 

These are the issues raised in a lawsuit between Peloton Interactive, Inc. on the one hand, and a group of music publishers and the National Music Publishers Association, Inc. (NMPA) on the other.

Peloton and Music Licensing. Peloton sells high-end, in-home stationary bicycles. An important feature of Peloton’s service is music-backed, instructor-led workout classes streamed to users via a built-in video screen. Some of these classes are broadcast live, and many are recorded and accessed on-demand.

Peloton doesn’t own its music, instead Peloton instructors create their playlists using popular recordings drawn from Peloton’s commercial music library.

To do this Peloton needs several music licenses. To play the sound recordings it needs master licenses from record companies. For public performance of the musical works (the compositions) it needs licenses from the performance rights organizations, ASCAP, BMI and SESAC. And, to reproduce the musical works along with videos, it needs synchronization licenses, so-called “sync licenses,” from the owners of the musical works.1

A sync license does not fall under any of the compulsory license provisions of the Copyright Act, and therefore sync licenses must be negotiated directly with the publishing companies that hold the copyrights. Here, it seems, is where Peloton’s attempt to acquire music rights fell short – Peloton was able to purchase blanket or “catalog-wide” sync licenses from what it calls the “major” publishers2 and many (but not all) “independent” publishers. 

The Publishers’ Copyright Suit. In March 2019 nine independent publishers filed suit against Peloton, claiming copyright infringement of more than a thousand musical works and seeking total damages as high as $150 million. According to the publishers Peloton transmitted these compositions with exercise videos without a sync license.

Based on the complaint the publishers appear to have a strong case, and it is unlikely that their core allegation – that Peloton was using these songs in videos without sync licenses – was false. If true, Peloton is looking at a significant settlement (or worse, a large court judgment). And, Peloton didn’t assert a credible public defense to the suit – in fact, it reacted by withdrawing the songs at issue from the library of tunes available to its instructors.

But it’s often the case that the best defense is a good offense, and Peloton has attempted to strike back in its countersuit.

Peloton’s Antitrust Counterclaim. In April Peloton filed a counterclaim against the nine publishers and NMPA, which it added as a defendant to its counterclaim. The heart of its claim is that, under the influence and urging of NMPA, the nine publishers refused to negotiate directly and individually with Peloton. Rather, they engaged in price fixing and a group boycott by insisting that Peloton negotiate licenses with NMPA on behalf of all the publishers. Peloton’s counterclaim states:

NMPA has instigated a coordinated effort with the Counterclaim Defendant music publishers to fix prices and to engage in a concerted refusal to deal with Peloton. Through these actions, NMPA has exceeded the bounds of legitimate conduct for a trade association and become the ringleader of concerted activity among would-be competitor music publishers, all in violation of the antitrust laws. . . . NMPA first sought to extract supracompetitive license terms from Peloton by negotiating collectively on behalf of a large (though unidentified) number of member publishers. . . . When NMPA’s collective negotiations with Peloton later stalled (for reasons NMPA never disclosed to Peloton), Peloton pursued direct negotiations with a number of music publishers. After participating in seemingly meaningful negotiations, however, several of the Coordinating Publishers suddenly — and virtually at the same time — cut off their negotiations and collectively refused to deal with Peloton. This refusal to deal with Peloton was at the urging of NMPA … 

Does Peloton have a viable antitrust claim? It’s a close call.

Buying Groups and Selling Groups. To understand the antitrust issues here I start with the concept of buying groups, or group purchasing organizations. The courts have recognized that most buying groups are economically efficient and (for example), enable small retailers to compete more effectively with larger retailers. Consequently, buying groups are “not a form of concerted activity characteristically likely to result in predominantly anticompetitive effects.” Northwest Wholesale Stationers v. Pac. Stationery & Printing Co. (USSC 1985). So long as the members of the buying group have a market share of less than 40% of buyers, they are protected by an FTC and court-approved safe harbor from antitrust liability for group negotiation of “input prices.”

However, the same is not true of “selling groups” – competitors who agree to set “output prices” at which they will sell or license their products. Output pricing by members of a trade association can easily cross the line over into price fixing, a per se violation under the antitrust laws.

“Hub-and-Spokes” Conspiracy. And that is what Peloton alleges occurred here. The antitrust theory behind Peloton’s counterclaim appears to be based on a “hub-and-spokes” conspiracy, similar to United States v. Apple (2nd Cir. 2015), where Apple acted as the organizer of an illegal horizontal price-fixing conspiracy among publishers in the e-book market.3 Peloton alleges that NMPA has acted as a “ringleader” similar to the role played by Apple in that case.

Twombly “Plausibility.” However, the NMPA and the publishers have already informed the court that they intend to file a motion to dismiss Peloton’s counterclaim. This motion is likely to rely on the argument that the counterclaim fails to allege facts that satisfy the Twombly “plausibility” requirement required for pleading an antitrust conspiracy.4 The publishers are likely to argue that the complaint contains insufficient allegations of concerted action between the publishers to support the inference that the publishers engaged in a horizontal conspiracy. In other words, in the language of the hub-and-spoke metaphor, there is no “rim.”

Indeed, Peloton’s antitrust claim provides no direct evidence of an overarching conspiracy between the publishers. There are no allegations of communications between the publishers from which the court could infer an agreement or conspiracy between them; there are no statements by informants or witnesses that support a horizontal conspiracy; and there are no specific allegations of back-room conspiratorial meetings between the publishers. Peloton alleges that several publishers pulled out of direct negotiations with Peloton at about the same time, but even these allegations of parallel conduct appear lacking: of the nine publishers alleged to be part of the conspiracy, Peloton provides specific allegations of parallel conduct as to only three.5

On this record Peloton may be hard-pressed to meet the “plausibility” standard and persuade the court that the conduct at issue is not the result of independent decision making on the part of the publishers – that is, that they would be better off having NMPA negotiate with Peloton on their behalf then negotiating directly.

Consequences if Peloton Prevails. That said, court rulings on motions to dismiss are unpredictable, and unappealable by a losing defendant. If Peloton’s counterclaim does survive a motion to dismiss its antitrust case represents a real threat to NMPA and the publishers. Any damages Peloton can prove will be subject to trebling, and NMPA and the publishers may be liable for Peloton’s attorney’s fees, which in an antitrust case can be substantial. And, the publishers’ copyrights may be unenforceable against Peloton based on the doctrine of copyright misuse.6

To add to this, there are practical issues that are not immediately apparent from the court filings. Can Peloton’s law firm represent all ten defendants (NMPA and the nine publishers) against this counterclaim? Or, are there sufficiently different interests that will require the publishers to each retain separate counsel, increasing their overall defense costs?

If the publishers’ lawyers are representing the publishers on a contingent fee basis, did they agree that this would include the defense of any counterclaims? If not, are the publishers (many of which are relatively small entities) prepared to incur the substantial cost of defending a complex, multi-party antitrust suit? Are they prepared, on behalf of their artist clients, to risk losing the right to enforce their copyrights against Peloton under the copyright misuse doctrine?

I would venture that the answers to most of these questions is “no.” Quite likely, once the potential impact of Peloton’s counterclaim sinks in, and assuming the counterclaim survives the motion to dismiss, some of the publishers will negotiate directly with Peloton and settle their copyright claims in exchange for a release from Peloton’s antitrust claims, leaving fewer and fewer of the “starting ten” counterclaim defendants to fight with Peloton. As the number of publishers left in the case decreases, the pressure on the remaining publishers will increase, eventually forcing them to the settlement table.7 How NMPA, the alleged “ringleader” – which owns none of the music involved – will navigate this minefield and emerge from this litigation is an open question.

Update: The music publishers have filed a proposed amended complaint greatly increasing the number of compositions at issue – and therefore the damages they are seeking from Peloton. Link here.

Update: This case was settled after the court dismissed Peloton’s antitrust counterclaim. (Link)


Ed Sheeran Should Settle “Lets Get It On”

Ed Sheeran Should Settle “Lets Get It On”

It looks like the music copyright world is facing another high-profile infringement trial. This time the songs at issue are Marvin Gaye’s 1973 “Get Lets it On1 and Ed Sheeran’s 2014 “Thinking Out Loud.” On January 2, 2019, a federal judge denied Sheeran’s motion for summary judgment in this case. Absent a settlement, the case will proceed to trial in federal court in New York.

This case raises some of the same questions that were at issue in the two cases decided recently by the Ninth Circuit – the Blurred Lines case (where the Ninth Circuit upheld a jury finding of infringement of another Marvin Gaye song), and the Led Zeppelin/Spirit Stairway to Heaven case, where the Ninth Circuit reversed a judgment for Led Zeppelin and remanded the case for a second trial. See Blurred Lines At The Ninth Circuit and Led Zeppelin, Spirit and a Bustle at the Ninth Circuit.

I discuss a few of the interesting aspects of the Get It On case below, but the bottom line is that Sheeran should settle rather than take the risk of proceeding to trial.

Admissibility of pre-1978 Sound Recordings. One issue present in the Ninth Circuit cases and here as well is the admissibility of a pre-1978 sound recording, in this case Get It On, which was recorded in 1973. The plaintiffs have rights to the musical work (the composition), but they do not claim a copyright in the sound recording. However, the plaintiffs argue that under the 1909 Copyright Act (in effect in 1973) a musical composition’s copyright extends to it first mechanical reproduction – in other words the 1973 sound recording of Get It On. Sheeran disagrees.

The judge dodged this arcane issue, leaving it for a later stage of the case. He held that his ruling on Sheeran’s summary judgment motion would be the same whether or not the copyright extends to the sound recording. Given the Ninth Circuit’s ruling in the Stairway to Heaven litigation,2 it’s unlikely the plaintiffs will be able to use the 1973 sound recording to prove infringement. This would leave them in the awkward position that faced the plaintiffs in Blurred Lines and Stairway, where the juries did not hear the original recordings of the plaintiffs’ songs, and the plaintiffs were left with no choice but to have a witness perform the sheet music in court. While this is a positive for Sheeran, it’s not enough to overcome the negatives in this case, especially given his live performance of the two songs, as I discuss below.

Commonplace Musical Elements. With respect to the substantive merits of the case, Judge Stanton rightly noted that many elements of Get It On are unprotectable. However, the parties and their musicologist experts disagreed on whether the chord progressions and harmonic rhythms present in the two songs are unprotectable as commonplace musical elements. This, the court held, is a factual question to be determined at trial.

The problem that Sheeran faces is copyright law’s recognition that a musical work may receive copyright protection based on a combination of commonly used and public domain elements. As the Ninth Circuit stated in its Stairway decision, “there can be copyright protection on the basis of a sufficiently original combination of otherwise non-protectable music elements. . . . the selection and arrangement of public domain elements could be considered original.” This is Second Circuit law as well – a work “may be copyrightable even though it is entirely a compilation of unprotectable elements.”(Knitwaves, Inc. v. Lollytogs (2nd Cir. 1995)).

At the heart of the infringement claim is Get It On’s “I-III-IV-V” chord progression and its distinctive harmonic rhythm, where the second and fourth chords are “anticipated” or placed ahead of the beat. Sheeran argued that this progression/rhythm is a common musical technique, but he has been able to point to only one pre-Get It On song that shared this chord progression and rhythm. While it’s unusual for a music copyright case to be based on a chord progression, it’s not unheard of, and the chord progression/rhythm in this case appears to be sufficiently original to avoid disqualification based on the scènes à faire doctrine3 and serve as the basis for a copyright claim.

The Substantial Similarity Test in the Second Circuit. A third factor that is likely to weigh against Sheeran at trial is the Second Circuit’s notoriously confusing test for substantial similarity in cases where the plaintiff’s work has both protectible and unprotectible elements, as is the case here.4.

In cases involving protectible and unprotective elements the court must apply the “more discerning” test and “attempt to extract the unprotectible elements5 from consideration and ask whether the protectible elements, standing alone, are substantially similar.” (Knitwaves, supra). Complicating what can be a challenging task even for trained musicians is the Second Circuit’s seemingly contradictory instruction that “we have disavowed any notion that we are required to dissect the works into their separate components and compare only those which are in themselves copyrightable.” Rather, the court (and presumably the jury) is “principally guided by comparing the contested design’s total concept and overall feel with that of the allegedly infringing work . . . as instructed by our good eyes and common sense.”

In my experience, jurors are easily confused by substantial similarity instructions in copyright cases,6 and the state of the law in the Second Circuit is likely to make that problem even worse than usual. Add to that the risk that some (perhaps all) of the jurors may be musically unsophisticated and the trial of a case like this becomes even more of a coin toss than most jury trials. The case may be vulnerable to the opinion of one or two jurors who have a musical background or be familiar with the musical genres at issue in the case.

The Sheeran Video. The capstone of the plaintiffs’ case against Sheeran is the unusual fact that Sheeran performed Thinking Out Loud in concert, during which he transitioned back and forth between Thinking Out Loud and Get It On (4:30 in the video). This is likely a first in a music copyright case.

While the judge may not permit the jury to listen to the 1973 recording of Get It On, he will be hard-pressed to preclude the plaintiffs from playing this video at trial. The plaintiffs have a near-insurmountable argument that this is an admission that the two songs are substantially similar. In fact, the judge refers to this video in his decision, suggesting that the video will be admitted at trial.7 That may be all the jury needs to conclude that the two works have the same “total concept and feel” or “aesthetic appeal” (another substantial similarity copyright standard in the Second Circuit).

Sheeran Should Settle. I’m well aware that many trained musicians (and amateurs) have weighed-in on Sheeran’s side of this case, and they may well be right. However, it’s unlikely Adam Neely or Rick Beato, or any knowledgeable musician, will end up on the jury in this case. Sheeran is more likely to end up with what one online wag described as a tone deaf judge and joe sixpack jury.

Assuming the plaintiffs are not grossly unreasonable, Sheeran would be better off settling this case, rather than risking an unfavorable outcome at trial that is likely to cost him more money than a settlement and damage his reputation.8

Sheeran seems to know he’s in trouble. After the judge denied his motion for summary judgment he took the unusual step of asking the judge to rehear the motion or, alternatively, certify an interlocutory appeal to the Second Circuit, a motion the judge promptly denied. This motion had virtually no chance of success, and served only to signal Sheeran’s desperation as the case proceeds toward trial.

Trial is scheduled for September 11, 2019.

[Update, July 2019]: The judge is delaying the trial of this case pending the Ninth Circuit’s en banc decision in Skidmore v. Led Zeppelin. That decision will probably not arrive until Spring 2020, so the trial is likely delayed by at least a year. Court order here.

Redigi – World’s First Used Digital Marketplace – Fails “First Sale” at Second Circuit

Redigi – World’s First Used Digital Marketplace – Fails “First Sale” at Second Circuit

I first posted on Capitol Records v. Redigi in March 2012 (Redigi Case Poses A Novel Copyright Question on the Resale of Digital Audio Files – Is “Digital First Sale Legal? Link), and posted a number of follow-up articles on this interesting case.1 Absent an appeal to the Supreme Court this long-running copyright case has finally come to an end with the Second Circuit’s December 12, 2018 decision holding that Redigi infringed the exclusive copyright right of reproduction with respect to the “second-hand” digital music files it sold via the Redigi system.

To understand this case it’s important to appreciate how Redigi’s system works. I explained this in detail in the post linked above, and the Second Circuit opinion describes it quite thoroughly as well. In short, Redigi acts as a broker for music files purchased and downloaded from iTunes. Redigi uploads a seller’s  music file to its own server and offers it for sale, deleting it from the seller’s computer, although the seller can continue to stream the file until it is sold. When a buyer selects it for purchase, it is retitled in the name of the buyer, and the seller loses access to it. The buyer may then stream or download the file to her computer or device.2

Redigi’s service irritated the record companies no end, and they sued for copyright infringement, asserting that Redigi was engaging in copyright infringement. Redigi, relying on the “first sale” doctrine, argued it did not.

The copyright first sale doctrine is an important exception to the copyright exclusive right of “distribution.” It allows the owner of a copyrighted work to sell the copy or phonorecord in which the work is fixed. This explains the existence of markets for second-hand books, records and CDs.3

Redigi argued that its service fell within the protection of first sale. The record companies argued that this analogy was inapt, since Redigi was not distributing the original file, but reproducing it on its server and on the buyer’s computer.4

The federal district court ruled in favor of the record companies (decision here5 and the Second Circuit (in an opinion written by Judge Pierre Leval, the Second Circuit’s prolific and influential copyright judge) agreed, reasoning as follows:

In the course of transferring a digital music file from an original purchaser’s computer, through ReDigi, to a new purchaser, the digital file is first received and stored on ReDigi’s server and then, at the new purchaser’s option, may also be subsequently received and stored on the new purchaser’s device. At each of these steps, the digital file is fixed in a new material object . . . The fixing of the digital file in ReDigi’s server, as well as in the new purchaser’s device, creates a new phonorecord, which is a reproduction . . . ReDigi’s server and the resale purchaser’s device on which the digital music files are fixed constitute or contain new phonorecords under the statute. (Emphasis added)

Redigi also argued (half-heartedly, it seems) that its system was protected by fair use, but this was an obvious loser. First, Redigi cannot show that it’s system is transformative. Second, Redigi makes identical copies of the whole copyrighted sound recording, which cuts against fair use. Third, the reproductions created by Redigi are sold in competition with the market for the original sound recordings, another negative factor.6 Each of these factors weighed against fair use, and Redigi lost on its fair use defense as well.

The bottom line: Redigi is enjoined from operating its service, and the company and its founders7 are on the hook for $3.5 million.

In an interesting postscript to this case, Redigi has developed a new methodology (“Redigi 2.0”) which allows a user to place a music file in the Redigi cloud server in the first place (it’s never downloaded to the user’s computer) and then simply transfer ownership to that file. Under this system Redigi never makes a copy (or enables users to make a copy), so it may not infringe the reproduction right. However, as part of a stipulated injunction in the district court Redigi agreed not to implement Redigi 2.0, and therefore its unclear whether the legality of this system will ever be tested in the courts.

Here’s my non-exhaustive take on how digital music files are treated under copyright law post-Redigi:

  • You purchase a CD that contains a digital music file authorized for sale by the copyright owner. You may sell it under first sale.8
  • You legally download a copyrighted music file to your computer and you transfer it to your smart phone for your personal use. This form of “space-shifting” is permitted based on fair use. Capital Record’s lawyer conceded this during oral argument before the district court in Redigi, and the Second Circuit commented on it favorably (in dicta) in its decision.9 It’s worth noting that the record companies have never sued a consumer for space-shifting legally acquired music files for personal use.
  • You legally download a music file to your computer and then upload it to a cloud service so you can stream it on your smart phone or speaker (e.g., an Amazon Echo). This is permitted based on fair use.
  • You purchase a device preloaded with music files authorized by the copyright owner. You can sell the device based on first sale, since this is a distribution, not a reproduction.
  • You download music files to your computer and sell your computer with your files on it. This is protected by first sale.
  • You download copyright-protected music files to your computer, transfer them to a thumb drive and delete them from your computer (i.e. “copy and delete”). You then sell the thumb drive. This is not protected by either first sale or fair use – based on Redigi this is an illegal reproduction.10
  • You legally download a music file to your computer, upload it to a cloud service, and then give your password to 25 of your closest friends or college dorm-mates so they can stream it. This is a violation of the copyright rights of reproduction, distribution and public performance. You lose.

Capitol Records LLC v. Redigi, Inc. (2nd Cir. Dec. 12, 2018)

Update: the Supreme Court denied review of this case, leaving the Second Circuit’s decision as the final word.