Privatizing Copyright

Xiyin Tang* March, 2023

Much has been written, and much is understood, about how and why digital platforms regulate free expression on the internet. Much less has been written—and even much less is understood—about how and why digital platforms regulate creative expression on the internet—expression that makes use of others’ copyrighted content. While § 512 of the Digital Millennium Copyright Act regulates user-generated content incorporating copyrighted works, just as § 230 of the Communications Decency Act regulates other user speech on the internet, it is, in fact, rarely used by the largest internet platforms—Facebook and YouTube. Instead, as this Article details, creative speech on those platforms is governed by a series of highly confidential licensing agreements with large copyright holders.

Yet despite the dominance of private contracting in ordering how millions of pieces of digital content are made and distributed on a daily basis, little is known, and far less has been written, on just what the new rules governing creative expression are. This is, in fact, by design: these license agreements contain strict confidentiality clauses that prohibit public disclosure of any and all of their contents. This Article, however, pieces together clues from publicly available court filings, news reports, and leaked documents. The picture it reveals is a world where the substantive law of copyright is being quietly rewritten. Agreements between digital platforms and rightsholders remove the First Amendment safeguard of fair use, insert a new moral right for works previously deemed ineligible for moral rights protection, and use other small provisions to influence and reshape administrative, common, and statutory copyright law. Further still, recent changes or lobbied-for changes to copyright’s public law seek to either enshrine the primacy of such private governance or altogether remove copyright rulemaking processes from government oversight, cementing the legitimacy of the new private governors.

Changing copyright’s public law to enshrine the primacy of such private governance insulates the new rules of copyright from the democratic process, transforming public participation in, and public oversight of, the laws that shape our daily lives. Creative expression on the internet now finds itself at a curious precipice: there is a seeming glut of low-cost or free content, much of it created directly by and distributed to users—yet increasingly regulated by an opaque network of rules created by a select few private parties. An understanding of the internet’s democratizing potential for creativity is incomplete without a concomitant understanding of how the new private rules of copyright may shape, and harm, that creativity.

Introduction

In recent years, a growing body of literature has turned public attention to the role that online intermediaries play in regulating free speech. These digital platforms have been described in internet law scholarship as “nonstate regulators” of the public sphere,¹ as private law enforcers acting within the “shadow[s]” of the state,² and as “new gatekeepers” controlling free expression.³ Indeed, the role that large internet platforms like Facebook play in regulating free expression looms so outsized in the literature that it has even, according to some, rendered the old dyadic model of speech regulation—in which the nation-state regulates the speech of those living within its borders—obsolete.⁴ Instead, as Professor Jack Balkin argues, “freedom of speech increasingly depends on a third group of players: a privately owned infrastructure of digital communication composed of firms that support and govern the digital public sphere that people use to communicate.”⁵ This digital infrastructure, Balkin has argued, “is important, if not crucial, to people’s practical ability to speak.”⁶

Yet there is, as this Article details, a fourth dimension to Professor Balkin’s triadic model of speech regulation: copyright owners who control not the infrastructure of the communication but much of its contents. That is, large swaths of user speech on digital platforms are not wholly original to the speaker. Instead, they incorporate bits and pieces of others’ copyrighted content: audio or audiovisual content that is owned not by the speaker, and not by the digital platform, but instead by large copyright holders like Disney, Warner Music, or Sony. How and why platforms moderate this type of user expression differs markedly, and hence requires a very different analytical framework, from traditional content moderation frameworks that have been well explored in the literature.⁷

Understanding this little-explored intersection of copyright and digital speech has important ramifications for the explosion of so-called “platform law” literature. ⁸ Rudimentary lay beliefs that digital platforms will block any user speech incorporating copyrighted content, for example, have led police to play Disney songs in an attempt to keep citizen-deployed accountability videos off social media. ⁹ Similarly, even scholars studying digital platforms mostly assume that copyright content moderation occurs under a statutory framework (§ 512 of the Digital Millennium Copyright Act ¹⁰), which allows for the “mass and easy removal of allegedly infringing copyrighted content, ending up with a significant chilling effect to freedom of speech.”¹¹ But neither of these accounts is wholly accurate. User-generated content containing copyrighted works is neither completely blocked nor removed en masse pursuant to a statutory framework. Instead, as this Article details, user speech that incorporates copyrighted content on large digital platforms like Google¹² and Facebook¹³ is governed by a series of highly confidential, private licensing agreements entered into between platforms and large copyright holders. Unlike the copyright statute, the use of private contracting gives copyright owners, in concert with platforms, power to create new rules that govern how millions of people share copyrighted content—without any need to resort to the legislative process.

This system of privatized copyright—in which large copyright holders leverage the power of platforms to enforce a preferred set of copyright policies that are passed down to the platform’s users through expansive platform terms of service¹⁴—presents a new and different example of the phenomenon of platforms acting as quasi-state actors.¹⁵ But the new private copyright has ramifications beyond the content moderation literature as well. For copyright scholars, insights into the new private copyright suggest that substantive, public copyright law¹⁶—such as the oft-repeated mantra that the United States does not recognize moral rights outside of the fine arts¹⁷ or that fair use is the most important First Amendment safeguard in copyright law¹⁸—matters little in the online sphere. By piecing together clues from publicly available news reporting and digital platforms’ own internal and external documents, including from one particularly pertinent leaked agreement that was entered into between Facebook and copyright holders, this Article instead reveals a world where the substantive law of copyright is being quietly rewritten and reshaped. Agreements between digital platforms and rightsholders remove the First Amendment safeguard of fair use, insert a new moral right for works previously deemed ineligible for moral rights protection, and use other small provisions to influence and reshape administrative, common, and statutory copyright law.

And finally, for contract law scholars and those who have long studied how private parties may contract around the substantive law,¹⁹ the new private copyright poses a fascinating question: how should we think about private contracts that govern the activities and behaviors of millions of nonparties to the original agreement²⁰—nonparties that are forbidden, by nature of the confidentiality clauses, from reading its terms? The very scale and reach of digital platforms as fundamentally public fora challenge traditional justifications for private ordering as “generally affect[ing] only their parties.”²¹ As Judge Easterbrook put it in an early case vindicating the use of “shrinkwrap” agreements for computer software: “[S]trangers may do as they please, so contracts do not create ‘exclusive rights.’”²² Instead, the rights that copyright holders have obtained through contracts with powerful digital intermediaries are beginning to look precisely like the exclusive rights created by the Copyright Act—applying to millions of strangers who have never seen, or even know, that such contracts exist.²³

This Article proceeds in four parts. Part I outlines the statutory framework that governs the distribution of user-generated content on the internet. Commonly referred to as the “safe harbor,” § 512 of the Digital Millennium Copyright Act establishes what’s known as a notice-and-takedown system, in which copyright holders can request the removal of user content upon a good faith belief that such use is not permitted by statutory copyright law.²⁴ The content moderation scholarship that has touched on copyright issues has focused overwhelmingly on § 512, with some describing it as a “federally mandated procedural system” that puts platforms in the role of copyright adjudicators.²⁵

Yet even as legal scholarship has continued to focus on § 512 as exemplary of copyright content moderation on the internet, the largest digital platforms—Google and Facebook—have been quietly moving away from reliance on the statute. Instead, as Part II details, copyright content moderation on those platforms is governed by a series of confidential private contracts that expressly dictate what users can and cannot do with copyrighted material. Part II details licensing agreement terms that create new substantive rights not present in the Copyright Act or that directly contravene existing statutory rights, as well as other privately negotiated terms that might even come to influence and reshape the common and administrative law of copyright.

While calls to reform platform content moderation practices outside of copyright have focused on more transparency and more restrictions on private power,²⁶ copyright law, on the other hand, has moved in much the opposite direction. As Part III discusses, the passage in Europe of Article 17 requires technology platforms to enter into private agreements and implement privately developed monitoring and filtering systems that automatically scan and remove certain user-generated content. While Article 17 may or may not serve as a model for similar changes to U.S. copyright law, the very global nature of internet platforms means European laws invariably become global laws as multinational companies voluntarily extend E.U. rules to govern their global operations.²⁷ Part III concludes with discussions of two other recent changes or lobbied-for changes in U.S. copyright law: one that removes all mentions of “public policy” in setting the compulsory licensing rates paid by digital streaming companies and another that would remove licensing negotiations between large music publishers and technology platforms from antitrust oversight by the Department of Justice, which had been overseeing such licensing activity since the 1940s.

Part IV concludes with some proposals for how to address the two biggest problems this Article has identified with regard to the privatized copyright regime: (1) a lack of transparency and (2) the supplanting of substantive law with private rules made without any input from those the rules govern. Part IV first argues that existing calls for reform, accountability, and greater transparency in content moderation practices are incomplete unless they sweep in copyright moderation practices as well. It then considers legislation that might make certain statutory rules of copyright, such as fair use, immutable rules that cannot be contracted around, similar to those enacted in Europe. And finally, it argues that nonparties to platform copyright contracts—users—can reinsert themselves in the process by bringing litigation to challenge the enforceability of certain contractual provisions. A celebration of the digital renaissance’s²⁸ abundance of creative content is incomplete without a concomitant understanding of how the new private rules of copyright are shaping, and may come to harm, that creativity.

I. The § 512 Statutory Scheme: Platforms as Private Courthouses

To understand how and why the largest internet platforms began entering into private agreements with copyright holders, it is important first to understand the statutory scheme that preceded it. Indeed, as discussed further below, certain quirks and inefficiencies in the statutory scheme were what led, in part, to platforms’ decisions to enter into private contracts and cease reliance on the statute entirely.

Whereas much of online expression is regulated (or insulated, as the case may be) by § 230 of the Communications Decency Act,²⁹ user expression that incorporates copyrighted content falls not under § 230’s purview but instead under the Digital Millennium Copyright Act (DMCA)’s safe harbor. Codified at 17 U.S.C. § 512, the copyright safe harbor is much like § 230 in that it insulates platforms like YouTube from copyright liability based on infringing content uploaded by its users, so long as certain conditions are met.³⁰ That is, § 512 sets forth a detailed “notice and takedown” procedure for resolving copyright owner claims of infringement based on user-uploaded content.³¹ Upon discovering infringing content on an online platform such as YouTube or Facebook, the copyright owner must first send a takedown notice to the platform, stating that the “use of the material in the manner complained of is not authorized by the copyright owner, its agent, or the law.”³² After receiving such a notice, the platform must expeditiously remove the infringing content in order to avoid exposing itself to infringement liability.³³

Takedowns are, in theory, subject to dispute by the user whose video was removed—an almost mini-adversarial, extra-judicial proceeding on the merits of the infringement claim.³⁴ Users can, for example, argue that the alleged infringing video in fact makes “fair use” of the copyrighted work. Perhaps the most well-known defense in copyright law,³⁵ fair use has been described as a built-in First Amendment safeguard,³⁶ allowing for uses of copyrighted works for purposes such as “criticism, comment, news reporting, teaching . . . , scholarship, or research.”³⁷ (When the New Yorker quotes from Sally Rooney’s Normal People in a profile on the author’s uncanny knack for documenting contemporary dialogue,³⁸ for example, this use would otherwise constitute copyright infringement were it not for the statutory right of fair use.)³⁹ Courts have held that § 512 requires a copyright holder to consider fair use prior to sending a takedown notification, because the statute requires all takedown notifications to include a “statement that the complaining party has a good faith belief that use of the material in the manner complained of is not authorized by the copyright owner, its agent, or the law.”⁴⁰ Thus, a “copyright holder who pays lip service to the consideration of fair use by claiming it formed a good faith belief when there is evidence to the contrary” is subject to liability under § 512.⁴¹

Nonetheless, in practice, rightsholders seldom consider fair use in demanding takedowns—and platforms likewise largely acquiesce to the thousands of takedown notices that they receive on a daily basis.⁴² And, despite the availability of the counter-notice procedure, a study conducted in 2016 by researchers at Berkeley Law School and Columbia University found that users, often legally unsophisticated, rarely took advantage of the provision.⁴³

Section 512 is by far the most well-understood—and, overwhelmingly, disliked—mechanism of copyright content moderation. Indeed, the platform law scholarship that has touched on copyright issues has focused almost exclusively on § 512.⁴⁴ Much of this scholarship uses § 512 as an example of the privatization of copyright law, as it delegates substantive infringement disputes that would normally be decided by a judge to digital platforms, who are incentivized to side with the copyright holders rather than the users for fear of exposing themselves to infringement liability.⁴⁵ Professor Rory Van Loo, for example, calls § 512 a “federally mandated procedural system” that puts platforms in the role of “copyright adjudicators.”⁴⁶ Van Loo argues that § 512 treats platforms as private courthouses, rendering default judgments against defendant users.⁴⁷

Other scholars, in line with the analogy to private judiciaries, argue that the adversarial process of notice-and-takedown suffers from an accountability deficit. Professors Maayan Perel and Niva Elkin-Koren write that “[r]emoving material that may qualify as fair use before notifying the alleged infringer and before giving her the opportunity to contest the removal in a hearing may result in ‘an extra-judicial temporary restraining order, based solely on the copyright holder’s allegation of copyright infringement.’”⁴⁸ They argue that § 512 fails to promote accountability in online copyright enforcement because the enormous number of takedowns on a daily basis renders it “impossible for the public as a whole to promptly identify and contest inappropriate content removals.”⁴⁹

If, by 2002, the notice-and-takedown process had largely been automated to meet the challenge of the sharing of millions of works incorporating copyrighted content on a daily basis, then it may seem only natural that private parties would soon begin evolving away from what many have argued was a broken statutory process to begin with.⁵⁰ Indeed, it’s not just scholars who dislike § 512—large digital platforms and copyright holders find little to like about the statute as well.

For copyright holders, because the statute puts the onus on the copyright holder, rather than the digital platform, to monitor for infringement, early user-generated sites like YouTube became a haven for infringing content.⁵¹ Copyright holders often likened the process of issuing takedown notices for copyright infringement to a game of whack-a-mole: as soon as one infringing video was removed, ten new ones would pop up in its place.⁵² By the time a consortium of copyright holders filed a putative class action lawsuit against YouTube in 2007, it was estimated that “75–80% of all YouTube streams contained copyright material.”⁵³

For digital platforms, the litigation that ensued over § 512 was certainly one reason they may have chosen to cease reliance on the statute and move into a world of private contracting instead. About a year before the class action litigation was officially filed against YouTube, the digital platform had already begun exploring the option of moving away from reliance on § 512.⁵⁴ Section 512, after all, creates a statutory exemption from infringement liability for digital platforms. But litigation to determine the exact contours of that exemption is expensive and time-consuming. A platform might prefer the certainty that comes with private contracting and a contractual release from infringement liability, instead. As discussed infra in Part II, the private agreements that arose in the wake of § 512 replaced a statutory public law scheme—with all its attendant litigation and uncertainty—with a privately determined set of rights and remedies that applied to content generated by users on the platforms.

Replacing uncertainty from litigation with certainty from contractually determined rights is likely one of the two main reasons digital platforms moved away from reliance on public law and into privately negotiated agreements. For example, Professor Matthew Sag posits that YouTube chose private contracting over reliance on the statutory safe harbor regime due to uncertainty generated by expensive and time-intensive rightsholder litigation over the contours of § 512’s application.⁵⁵ Other scholars have likewise pointed to litigation-related uncertainty as the main animating factor behind YouTube’s pivot from reliance on statutory protections to negotiated licenses.⁵⁶

Yet others have suggested that, in YouTube’s case, it arose specifically out of Google’s desire to monetize YouTube videos, as described above—that is, to sell targeted advertising to viewers based on their watch patterns.⁵⁷ Because the DMCA specifically requires that the technology platform have no knowledge of the infringement in order for the safe harbor to apply,⁵⁸ targeted advertising, which would require the platform to track user activity to better sell ads tailored to their behavior, would substantially weaken any argument that the platform lacked actual knowledge of infringing user activity.

Both of the above explanations converge under a common theme: while initially seemingly counterintuitive, technology platforms and copyright holders will often choose to enter into agreements that are more restrictive and more expensive than the statutory requirements precisely because they ultimately better serve both parties’ business objectives—many of them unapparent or highly technical to all but the most seasoned industry experts.⁵⁹ And at bottom, one explanation needs little specialized knowledge to ring true: dealmaking is better politics than adversarial litigation. As Professor Kristelia García puts it, a “copacetic working relationship can lead to content exclusives, more listeners, and more revenue for both parties.”⁶⁰ “Artist-friendly” is good PR.

If scholars liken the notice-and-takedown procedure of § 512 to one-sided adjudication, then the privatized copyright described in the following Part might be more akin to private legislation. Internet law scholars largely define the term “platform law” to encompass the privately developed set of rules—such as terms of service, community standards, content moderation practices and decisions, and internal guidance provided to employees—that digital intermediaries have developed to regulate user activity on their platforms.⁶¹ Because of the sheer scale of the largest digital platforms, their massive number of users, and the breadth of user activity that takes place on them daily, scholars have likened companies like Meta and Google to sovereign states—each developing unique “laws” to govern its myriad users.⁶² If § 512 in theory required copyright holders to consider whether a user’s video constituted fair use—but few did in practice—then perhaps, when it came time to supplant the statute with private agreements, fair use should be written out of the equation completely. As discussed below, that is precisely what copyright holders, working in concert with digital platforms, did.

II. The New Private Copyright Law: Platforms as Legislators

While a number of scholars have pointed out the adjudicatory role of platforms under § 512, the most powerful digital platforms today do not rely on the statute.⁶³ Instead, as this Part discusses, companies like Meta and Google have chosen to enter into a series of private, highly confidential agreements with copyright holders that determine how users can share content incorporating copyrighted works—what this Article refers to as a new private copyright law.

A. The Rise of Private Contracting

This Section details the development and rise of private contracting to moderate user-generated content containing copyrighted works—contracts that supplant the statutory § 512 practice. These private agreements that digital platforms and powerful content holders enter into create the web of rules that govern how nonparties to the agreements—the platforms’ millions of users—share, use, and distribute creative expression containing copyrighted content.

1. YouTube’s Content ID

In 2006, YouTube began exploring the option of moving away from reliance on § 512 by negotiating license agreements with several large content owners, including Time Warner, Disney, and EMI.⁶⁴ A year later, Content ID, “a proprietary, voluntary agreement between YouTube and a select group of content owners” that would make it such that YouTube no longer needed to rely on § 512, “was up and running.”⁶⁵

Under YouTube’s licensing agreements, copyright holders first submit to YouTube large datasets of audio-only or audiovisual content owned by them—called “reference files”—and accompanying metadata.⁶⁶ Content ID’s database has now expanded to include more than eighty million files of audio and visual content.⁶⁷ When a YouTube user uploads a video, the video is first scanned against Content ID’s digital database of reference files to determine if there is a match, and, if so, it applies one of three actions to it. The rightsholder specifies the action up front: (1) allow the video to remain up and for YouTube to run ads against it in exchange for splitting the ad revenue with the copyright holder (a practice known as “monetization”); (2) allow the video to remain up and track how users are engaging with the content; or (3) block it from YouTube altogether.⁶⁸ None of these three options is a right allotted to copyright holders under the DMCA,⁶⁹ which only provides copyright holders the ability to request the takedown of infringing videos.⁷⁰ Specifically, with regard to the third option available under Content ID, under the DMCA, a user firsts upload a video incorporating infringing content onto the site—only then would a rightsholder be able to request that it be taken down.⁷¹ Meanwhile, before YouTube processes the takedown notice, the content remains up for anyone to view. Content ID turns the DMCA on its head by allowing a rightsholder to block the video before it’s posted at all—what others have argued constitutes a potential private prior restraint of speech.⁷²

Nonetheless, publicly available statistics from YouTube’s Content ID program show that the vast majority of infringing content on YouTube is monetized rather than blocked.⁷³ Yet YouTube’s monetization program is the exception, not the norm. Even a company as well resourced as Meta, while having made some strides towards monetization, is nonetheless unable to monetize large swaths of user-generated content, as discussed in more depth below.

2. Meta’s Rights Manager

While YouTube’s process of entering agreements with content holders and its development of Content ID is by far the most well-known and oft-discussed example of a technology platform opting out of a statutory copyright regime in favor of private contracting,⁷⁴ it is far from the only one. Unsurprisingly, in the two decades following the enactment of the DMCA, more and more platforms hosting user-generated content shifted to some version of private rulemaking and dispute resolution rather than following the letter of the statute. One of the most notable examples is Meta (formerly Facebook)—which owns, among others, Facebook and Instagram.⁷⁵ Like YouTube, beginning in the late 2010s, Facebook began the process of moving away from the statutory § 512 scheme by choosing instead to enter into a slate of deals with major record labels, music publishers, and performing rights organizations.⁷⁶ In conjunction with these deals, Meta developed its own, similar version of Content ID, called Rights Manager.⁷⁷

Like YouTube’s Content ID, Rights Manager is an algorithmic copyright management tool that scans content uploaded to Facebook and Instagram and then runs it against a database of copyrighted content.⁷⁸ Using Rights Manager, publishers can “[e]asily upload and maintain a reference library of the video content they want to monitor and protect;” “[c]reate rules about how individual videos may be used;” “[i]dentify new matches against protected content;” “Whitelist specific Pages or profiles to allow them to use their copyrighted content;” and “[p]rotect their reference library at scale with the . . . Rights Manager API.”⁷⁹ In short, content owners can choose to block content, monetize content, have video content automatically credited to the owner, or report videos that the Rights Manager algorithm has flagged as potentially infringing. In September 2020, Rights Manager was expanded to include not just audio content but images as well, allowing for the digital fingerprinting of user-uploaded image content at scale.⁸⁰

Unlike YouTube’s monetization success story,⁸¹ monetization on Meta’s applications has long been problematic. To begin, consider how different YouTube is as a product compared to Facebook or Instagram. YouTube videos appear on a discrete page and are usually at least a few minutes long, allowing advertisers to place ads either at the beginning or in the middle of these longer-format videos.⁸² On the other hand, the vast majority of videos posted on Instagram are short—often under a minute⁸³—appearing in quick succession as a user views them not on a discrete page but rather as part of a continuous social feed. As one industry commentator puts it: “[M]onetizing short-form content remains problematic, because you can’t attribute pre or mid-roll ads to specific clips, like you can with longer posts.”⁸⁴ Currently, according to Meta’s website, only videos that contain so-called “in-stream ads” (ads that play either before or during a video) are eligible for monetization.⁸⁵ And because the company only places in-stream ads in videos three minutes or longer, and, on the Facebook platform, only from business pages, this excludes the vast majority of videos posted on the platform. The majority of videos posted on Instagram fall under the one-minute mark, and no video posted on what one would think of as a traditional Facebook page—a user profile—is eligible for monetization.⁸⁶ And indeed, Facebook’s 2018 agreement explicitly disclaims any monetization ability, providing payment not based on usage but rather on a single “non-recoupable lump sum” fee—a one-time payment to be made to rightsholders up front.⁸⁷

3. Other Platforms Follow Suit

While well-resourced technology giants like Meta and YouTube are able to build their own copyright management systems, other less-resourced internet platforms today have also chosen private contracting and private content management tools over reliance on the DMCA.⁸⁸ As just one example, SoundCloud—which, like early YouTube, once served as an anything-goes mecca for remixes and mash-ups, and as a platform for artists to bypass the lengthy and prohibitive process of obtaining rights for unauthorized samples or covers of other copyrighted works⁸⁹—began publicly (and proudly) striking deals with music labels and publishers.⁹⁰ As one news outlet lamented, the move signaled the end of “the SoundCloud [we] once knew—an audio playground filled with unexpected surprises and treasures”—to a conventional streaming service with an established, approved catalog of songs.⁹¹ Other platforms that have struggled with user and rightsholder complaints over repeated takedowns and alleged infringement, respectively, under the DMCA have also begun to explore private licensing avenues, touting such deals as “productive partnerships” with the music industry.⁹²

B. How the New Private Copyright Is Rewriting Substantive Law

How does copyright’s new private law—entered into behind closed doors between large, sophisticated business entities that are each driven by their own hidden motivations and agendas—differ markedly from statutory copyright? Of course, it is impossible to uncover, let alone detail, every single one of the divergences from publicly available law, as each of the agreements is subject to strict confidentiality clauses. But one particularly pertinent leaked agreement between Facebook and rightsholders is illustrative of the new private copyright—and just how expansive such confidentiality clauses can be.⁹³

The agreement, published by the industry website Digital Music News under the headline Here’s the Entire Facebook Contract for Music Publishers & Songwriters, was a standard form Facebook contract that it had, in this case, entered into with independent music publishers—sometimes called “indies.”⁹⁴ Despite the fact that the “CONFIDENTIAL” stamp clearly appears on every single page of the agreement, and notwithstanding the broad confidentiality provision on page 4, the agreement was likely leaked because of the sheer number of licensors at issue.⁹⁵

The confidentiality clause, on page 4 of the agreement, provides:

Each Party will keep the terms of this Agreement and the data provided or generated pursuant to this Agreement (including any reports provided pursuant to this Agreement) confidential . . . .
. . .
Without prejudice to the generality of the foregoing, Publisher, without the prior written approval of Facebook, may not publicize, in a press release or otherwise, the existence or terms of this Agreement or any other aspect of the relationship between the Parties.⁹⁶

That last excerpted sentence should give us pause. That is, not only are all of the agreement’s terms and any and all data generated pursuant to it confidential—such as, for example, any information about which, how many, or why videos were blocked or muted for rights violations—so too, is the mere fact that the agreement, or any contracting relationship between Facebook and rightsholders, exists at all.

And, as this Section details, these highly confidential agreements are quietly rewriting public copyright law.

1. Creating New Substantive Rights

It is often said that U.S. copyright law, unlike European copyright law, is rooted in economic, rather than personhood, justifications.⁹⁷ Indeed, the United States had long resisted joining the primary treaty governing international copyright law, the Berne Convention,⁹⁸ in part because of its expansive moral rights provisions,⁹⁹ which granted copyright owners wide-ranging rights to challenge modifications, distortions, and, in broad language appropriate of the personhood nature in which moral rights springs, “other derogatory action” to their works that are “prejudicial to [their] honor or reputation,” as well as a right of attribution.¹⁰⁰ When the United States finally ratified the Berne Convention in 1988,¹⁰¹ Congress enacted just a limited form of moral rights, choosing to limit it to visual works of art only—specifically, “a painting, drawing, print, or sculpture, existing in a single copy, [or] in a limited edition of 200 copies or fewer that are signed and consecutively numbered by the author.”¹⁰² Indeed, the legislative history surrounding the passage of the Visual Artists Rights Act of 1990 (VARA) specifically emphasizes the “limited application” of the Act to works of visual art only, noting that Congress had considered, but ultimately decided to reject, a more expansive moral right covering other types of copyrighted works.¹⁰³ Notably, while other industries had lobbied for the right to prevent the distribution of modified copyrighted works in manners that they objected to, Congress specifically chose to reject a broader moral right for copyrighted works beyond the narrowly defined set of visual artworks.¹⁰⁴

Courts have subsequently rejected moral rights–like claims brought by copyright owners not covered by VARA’s narrow ambit.¹⁰⁵ The Supreme Court, in its decision rejecting a right of attribution brought by the owner of an audiovisual work, noted that the moral right of attribution is “carefully limited and focused: [i]t attaches only to specified ‘work[s] of visual art,’ . . . is personal to the artist, . . . and endures only for the ‘life of the author.’”¹⁰⁶ Other courts have noted that no federal cause of action for moral rights exists, precisely because VARA “protects only authors of a work of visual art.”¹⁰⁷

But what copyright holders like music publishers or record labels did not receive in the statute, they are now receiving, at scale, with the new private copyright.

Specifically, a section titled “Blocking” on page 2 of the agreement between Facebook and rightsholders provides rightsholders with the right to request takedowns of user-generated content incorporating copyrighted works “due to a bona fide [song]writer objection.”¹⁰⁸ While the agreement purports to limit the music publisher to exercise its “Blocking Right in a manner that will [not] have more than a de minimis impact in quantity and quality of Publisher Compositions on the Facebook Properties,” publicly available data suggest that labels and publishers are, in fact, submitting thousands of takedown requests at the behest of their artists, per day.¹⁰⁹

Consider, for example, the testimony of the Eagles’ Don Henley before the Senate Judiciary Subcommittee on Intellectual Property in June of 2020—well after the music licensing agreement between Facebook and publishers was signed.¹¹⁰ Henley noted that his record label, Universal Music Group, which also handles music publishing rights for both the Eagles and for his solo catalog, has a team of

sixty people who sit in a room with computers, and all they do all day long, five days a week—sometimes six days a week—is deal with the platforms such as YouTube and Facebook. They file claims and they issue takedown notices for the Eagles and for myself. Those amount to between 200 and 500 claims a week.¹¹¹

While Henley is one of the most vocal opponents of making his music available on social media, he is certainly not the only artist issuing hundreds of takedown objections per week. It has been publicly reported that other artists who famously refuse to make their music available on the two largest user-generated video platforms, Facebook and YouTube,¹¹² include “Fleetwood Mac, AC/DC, Pink Floyd, Black Sabbath, [and] Ozzy Osbourne,” all of whom “regularly block videos that use any amount of their music, regardless of the context.”¹¹³

The artists that object to making their music available on large social media platforms might have very strong, visceral views about Facebook or Google. Roger Waters of Pink Floyd, for example, has been publicly reported as strongly (to put it lightly) criticizing Facebook and its founder, Mark Zuckerberg.¹¹⁴ For all these reasons, artists, as well as those scholars who have criticized the United States’ restrictive stance on moral rights, likely welcome the artists’ objection clause as a contractual expansion of moral rights to more types of artists and works.¹¹⁵ In this sense, the artists’ objection clause is doing exactly what proponents of moral rights argue we should be giving an artist: a greater opportunity to refuse uses of her work that deeply offend her.¹¹⁶

But, as the following Section discusses, in many instances, the artists’ objection clause does not just provide greater substantive rights than what the Copyright Act affords—it might also directly contravene existing substantive rights in the Copyright Act.

2. Contravening Existing Substantive Rights

The doctrine of fair use is considered so integral to copyright law that even earlier efforts at private self-governance, such as the Principles for User Generated Content Services (UGC Principles), explicitly provided for fair use accommodations. The UGC Principles were a set of industry guidelines for managing the uploading and distribution of user-generated content on the internet that had been voluntarily agreed to among several large internet companies and several large copyright holders in 2007. Unlike the new private copyright, the UGC Principles were made publicly available online and included, as one of their four core objectives, the “accommodation of fair use of copyrighted content on UGC Services.”¹¹⁷ The enumerated fifteen-paragraph principles that follow require copyright owners to consider fair use prior to sending takedown notices and require platforms that use filtering technology to implement such technology in a way that considers fair use.¹¹⁸ Accordingly, the UGC Principles were lauded by commenters as evidence of private ordering that complemented existing laws such as “the fair use doctrine.”¹¹⁹

Of course, in practice, it is notoriously difficult to accommodate fair use at the scale the new digital age of information sharing demands.¹²⁰ And so, perhaps because of this very difficulty, and in sharp contrast to the UGC Principles, nothing in the Facebook agreement with music publishers mentions fair use at all.¹²¹ And, in fact, the artists’ objection clause discussed in the previous Section may directly conflict with fair use doctrine.

Imagine, for example, that a video uses a few seconds of John Lennon’s song Imagine juxtaposed against images of a secular world torn apart by war. This was the precise use—but in the form of a full-length film, rather than in a social media use—that resulted in a filing of a copyright infringement suit by Lennon’s heirs.¹²² The court held that the use constituted fair use, because the purpose was “to criticize the song’s message,” and thus refused to enjoin the film from nationwide distribution (including theatrical release).¹²³ But if the distribution had occurred solely on Meta’s platforms rather than in theaters, Lennon’s heirs could simply exercise their artists’ objection right to remove the video from distribution. The artists’ objection need only be “bona fide”—it is not subject to any fair use carve-out.¹²⁴ (Whether the filmmakers would be able to bring a lawsuit arguing that the video was wrongfully removed, on the other hand, will be discussed infra in Section IV.B.)

Ultimately, there is a bigger problem in all of this, beyond just the point that the new private law of copyright directly contravenes statutory law by allowing rightsholders to block and remove videos that would otherwise constitute permissible fair uses. And that is that it does so with no public visibility, let alone accountability, into the removal and blocking process: freed of the need to send a specific DMCA notice pursuant to § 512, the new takedowns occur in the shadows, without fear of being entered into some public archive somewhere for academic researchers or reporters to scrutinize and critique. After all, the agreements have classified takedown data effectuated under the agreements as confidential information that must not be disclosed.¹²⁵ As a result, users have little idea who, or what, is even animating these decisions—whether it was removed because of some action on the part of a copyright holder rather than the platform’s own capriciousness and censorship, or whether it is a public statute such as the DMCA that allows for these removals, rather than private contracting.¹²⁶ Indeed, internet law scholars who regularly research and write on this very topic confuse just who is responsible for initiating takedowns for copyright infringement. They often attribute responsibility single-handedly to platform overreach rather than copyright holder overreach. Ironically, they also often attribute responsibility largely to the DMCA rather than private agreements—a fact that does not bode well for how much we can expect everyday users to understand the laws animating blocking decisions.¹²⁷

C. Private Copyright as Influencing and Shaping Public Law

This Part has so far discussed examples of how the new private copyright law may provide rights that Congress had specifically considered but declined to provide in the Copyright Act—rights which may simultaneously conflict with important substantive defenses that are present in the Copyright Act, like fair use.

But these privately negotiated agreements do more than simply provide greater substantive rights by contract (or route around statutory rules by contract). Instead, because certain doctrines—most notably, fair use—and administrative copyright proceedings specifically take industry practice into account, what parties privately contract for may also eventually affect the public, substantive law of copyright.

1. Private Copyright Influences and Changes Public Common Law

Fair use, as a common law doctrine developed through the accretion of case law, is notoriously susceptible to industry practices.¹²⁸ This is especially so because the critical fourth factor of the four-factor fair use test looks to “traditional, reasonable, or likely to be developed” licensing markets.¹²⁹ Professor James Gibson calls the practice of licensing within areas where the common law is still developing—in which judges are still deciding how best to allocate entitlements—“gray areas.”¹³⁰ The industry practice of licensing within these gray areas, however, “eventually makes those areas less gray, as the licensing itself becomes the proof that the entitlement covers the use.”¹³¹ Or, as Professors Christina Bohannan and Herbert Hovenkamp bluntly put it: “[T]he existence of [a] license[] tends to reify the notion that the right to control such uses exists.”¹³² As discussed below, two such “gray areas” that the new privatized copyright is skewing ever less gray are (1) commercial uses and (2) reproductive uses.

a. Commercial Uses as Infringing

One such gray area in copyright law has long been commercial use—uses of copyrighted works that happen to make money for the secondary user. While it may now seem noncontroversial that works sold for profit—such as a movie or an album—could still be eligible for the fair use defense, this was not always a foregone, or obvious, conclusion. The statutory text merely provides that courts shall consider “whether [the secondary use] is of a commercial nature” in its evaluation of the first factor.¹³³ In two successive fair use decisions, Sony Corp. of America v. Universal City Studios, Inc. and Harper & Row, Publishers, Inc. v. Nation Enterprises, the Supreme Court repeatedly stated that “every commercial use of copyrighted material is presumptively an unfair exploitation of the monopoly privilege that belongs to the owner of the copyright.”¹³⁴ Yet almost a decade later, the Court changed tack. In Campbell v. Acuff-Rose Music, Inc., it specifically overturned the lower appellate court’s holding that gave “virtually dispositive weight to the commercial nature” of the infringing song at issue, holding instead that the “commercial or nonprofit educational purpose of a work is only one element of the first factor enquiry into its purpose and character.”¹³⁵

Yet Facebook’s agreement with rightsholders does something different. It gives dispositive weight to the commercial nature of the use, holding that any “commercial use” constitutes a presumptive “unauthorized” use.¹³⁶ Commercial uses—defined in the agreement as videos “uploaded by business page (on Facebook) or business account (on Instagram) holders (‘Business Users’)”—are funneled into a separate so-called “commercial content review queue,” with the explicit purpose of “identifying to [the copyright holder] potential licensing opportunities.”¹³⁷ So, for example, a New York Times business page that wished to post a critique of a musician’s latest album by using a snippet of the song would be engaged in an “unauthorized” use, as determined by the agreement. If it wishes to use copyrighted music, it must obtain a license.

The Times’s decision to obtain a license might seem inconsequential and uncontroversial, in and of itself. But combine enough of these isolated instances of obtaining licenses for for-profit uses, and what you have is private practice influencing public, substantive law. Courts making fair use determinations have expressly found that the existence of an available licensing market, as well as a plaintiff’s successful licensing practice, makes a use less likely to be fair.¹³⁸ If a music publisher can show, for example, that it has always been able to receive licensing revenue for uses of its music by magazines or newspapers on digital platforms regardless of the content, a court might be persuaded that such uses must always be licensed, shifting fair use doctrine as a whole back towards an emphasis on commerciality. Professor Gibson calls this phenomenon “doctrinal feedback,” in which the “aggregate effect” of isolated, “unobjectionable, even laudable” decisions to obtain licenses results in “an expansion in the reach of intellectual property rights—an expansion completely unconnected to lobbying successes and courtroom victories.”¹³⁹

In a further testament to the opaque nature of the new private copyright, Facebook itself has never confirmed the limitation on business account posts in any of its public-facing communications. Instead, confused users are left to speculate amongst themselves, in internet forums, through Instagram direct messaging, and cries for help to other Instagram users, on if, and why, business pages are blocked from accessing music.¹⁴⁰ And this is not the only instance in which the confidentiality clauses governing the new private copyright have resulted in few clear user guidelines for creative speech. As the following Section discusses, the COVID-19 pandemic unexpectedly created a rare public outcry—and resulting attempt at clarification—over Facebook’s copyright policies.

b. Reproductive Uses as Infringing

As individuals found themselves staying home during the start of the COVID-19 pandemic and ensuing lockdowns, a rare bright spot emerged. Churches that had formerly held services in person started live streaming them instead, using Facebook Live.¹⁴¹ Isolated friends joined dance parties on Instagram Live.¹⁴² Fireside chats, too, moved onto Instagram Live.¹⁴³ Indeed, the formerly embattled Live product, which Facebook had mostly sidelined for lack of popularity, saw an astounding spike in usage as quarantined individuals everywhere convened in virtual rooms.¹⁴⁴

The surge in popularity also brought to light, to many for the first time, the new private copyright. Church organizations found their prayer services muted—and soon realized it was because the songs that played during those virtual services were copyrighted.¹⁴⁵ DJs who invited their fans for a virtual set found their live streams suddenly blocked and muted.¹⁴⁶ Notably, in a testament to just how obscured the new private law of copyright is, several news outlets reported that Facebook was introducing “new” music terms of service starting October 1, 2020.¹⁴⁷ Other blog posts, however, felt the need to explain to their audiences that while “[t]he internet is buzzing with the ‘new’ restrictions on music embedded in videos,” the “truth is that Facebook has had these music guidelines in force since 2018.”¹⁴⁸ It just took a global pandemic and lockdown to surface it. In a rare public post, Facebook addressed the situation:

Music is a bonding force in normal times. During difficult and isolating times like this, we know it can be even more important. As social distancing has forced everyone to stay apart, more people have turned to Instagram and Facebook Live to stay connected with their communities. This rapid rise in usage has created a lot of good during this crisis—raising money for frontline workers and underserved communities, driving awareness of healthy habits, encouraging people to stay safe by staying home, and bringing people together through new forms of entertainment. But it’s also highlighted some confusion across the community—especially around the use of recorded music in Live on both Facebook and Instagram.¹⁴⁹

Yet, the company stated, because “the specifics of our licensing agreements are confidential,” it could only “share[] some general guidelines to help you plan your videos better.”¹⁵⁰ The guidelines, however, did not set out clear parameters for when and why a live stream would be blocked for a copyright violation. Instead, they only provided vague guidance, stating, for example, that “[t]he greater the number of full-length recorded tracks in a video, the more likely it may be limited,” that “recorded audio should not be the primary purpose of the video,” and that “shorter clips of music are recommended.”¹⁵¹

But these guidelines created scant workable standards for communities hoping to use the popular social media platform during the pandemic. Churches were advised, for example, that due to the possibility of muting for copyright violations, it was “too risky to [only] stream to Facebook,” and religious worship sites argued that “[c]hurches should have alternate, more reliable sites host their services.”¹⁵²

The leaked Facebook agreement, however, begins to add some clarity to the publicly available guidelines. In a section titled “Specifications and Product Restrictions” on page 3 of the agreement, it is revealed that Facebook has agreed, “pursuant to commercial agreements with record label partners,” to “implement certain abuse prevention mechanisms to inhibit abusive use of User Videos as an audio-only music listening experience . . . as set forth in Exhibit B.”¹⁵³ Moving to Exhibit B, then, we can see that the “abuse prevention mecasures” Facebook will put in place include the blocking of “User Videos . . . that contain one or more nearly complete [songs] with a static image” and the blocking of “any . . . User Video . . . that contains five or more nearly complete [songs].”¹⁵⁴ These deal terms, then, help explain why Facebook publicly advised that “recorded audio should not be the primary purpose of the video” (because audio-only videos are not permitted under the agreement) and why Facebook advised that “[t]he greater the number of full-length recorded tracks in a video, the more likely it may be limited.”¹⁵⁵ The agreement tells us exactly how many tracks may appear in a video before it will be blocked: five.¹⁵⁶

The agreement, in essence, has taken a position against rote copying: reproductive copying that merely reproduces the original.¹⁵⁷ Whereas reproductive copying is deemed abuse under the agreement, leading to its automatic blocking, copying that transforms the original work—so-called “productive” uses—are permissible.¹⁵⁸ Yet the question of whether reproductive copying should in certain instances be permitted as fair use has long been an area of contested debate. Appropriation art, for example, true to its name, slavishly reproduces the original, and courts have vacillated between holding that such rote copying constitutes clear-cut infringement or else paradigmatic fair use.¹⁵⁹ Likewise, scholars have fought for the idea that pure copying, rote copying, and simple reproductive copying can serve important First Amendment values.¹⁶⁰ Certainly, the examples from the pandemic bolster the point that reproductive uses, even ones that utilize multiple copyrighted works, may be good candidates for arguing for the continued viability of fair use as applied to reproductive copying. A priest may wish to illustrate his sermon with a selection of songs that speak on hope, loss, and grief, even if the songs themselves are not otherwise transformed and even if there is no commentary upon the songs themselves. By deeming such a use de facto unauthorized, the new private copyright is, just as it did with commercial uses, closing more and more secondary uses off from any colorable claim to fairness.

While fair use decisions have always been less receptive to reproductive uses of copyrighted works,¹⁶¹ they may begin to swing even further in the direction of only permitting productive rather than reproductive uses. Just as with commercial uses that are subject to separate licensing negotiations under Facebook’s agreement, rightsholders may likewise point to the fact that reproductive uses have traditionally constituted unauthorized uses for which separate licenses are required, making market harm more likely and a claim to fair use, in turn, less likely.¹⁶²

2. Private Copyright Creates Public Copyright Regulations

While this Article focuses, for the most part, on user-generated content—content created by and for users, uploaded to sites like YouTube and Facebook—it is important to emphasize that this phenomenon is not just limited to user-generated platforms. All across the digital space, rightsholders are negotiating for substantive rights that will come to reshape what the public law of copyright looks like. And just as with the common law doctrine of fair use, administrative copyright proceedings, too, take industry practice into account. One example involving the definition of a “stream” in public copyright rulemaking illustrates how privately determined definitions in industry agreements have a role to play in creating substantive public copyright law.

It began with uproar from fans of the Korean pop group BTS over uncounted Spotify streams.¹⁶³ Following the release of its new single, the pop group’s fans launched a campaign to boost the song’s popularity by playing it on repeat, including on popular streaming platforms like Spotify.¹⁶⁴ Yet something odd happened with the streaming count for the BTS single Butter. Fans began noticing that not all streams were being counted.¹⁶⁵ Angry fans reached out to Spotify’s support team, which confirmed that the streaming service has systems in place to filter out what the platform considers to be illegitimate streams.¹⁶⁶ Fans, in turn, began demanding greater transparency over how Spotify filters out certain streams, giving rise to the hashtag #InvestigateSpotify.¹⁶⁷

Each stream on Spotify implicates two copyrighted works: a sound recording and a musical composition.¹⁶⁸ The former is licensed to Spotify pursuant to a series of confidential licensing deals. However, as discussed infra in more detail in Section III.B, streaming royalties for musical compositions are set by an administrative proceeding before the Copyright Royalty Board (CRB). Accordingly, public filings in those proceedings might provide one clue as to how Spotify tracks streams. As it turns out, those filings and the ultimate decision in the most recent rate-setting revealed that streaming services’ own internal policies eventually translated to substantive, public copyright regulations that now govern all digital streams of musical works.

In the streaming rate-setting proceeding for the 2018–2022 rate period, Apple’s music streaming service proposed, for the first time, an exemption for a “fraudulent stream,” which it defined as “a stream that a service reasonably and in good-faith determines to be fraudulent.”¹⁶⁹ Apple further proposed defining a royalty-bearing stream as any “play of a sound recording of a copyrighted work lasting 30 seconds or more,” thus omitting streams of under thirty seconds from being counted or tracked as a proper stream.¹⁷⁰ Notably, in discussing these requests to omit certain types of streams from the royalty-bearing calculus, the final opinion in the rate-setting proceeding states: “Apple contends that the [thirty-second] time threshold is a feature of [REDACTED].”¹⁷¹

Redactions in the public opinion are made for confidential information—which the Protective Order in the case defines as “commercial or financial information” that is competitively sensitive.¹⁷² Thus, the very fact that the information was redacted speaks volumes about Apple’s proposal. Apple’s proposal to limit royalty-bearing streams to nonfraudulent streams of thirty seconds or longer must have some basis in either how Apple actually tracked and accounted for streams, or in its confidential deals with record labels, in order to constitute commercially sensitive information warranting redaction. And realistically, it is the latter that drives the former: Apple almost certainly would not be able to avoid paying record labels for certain streams—that is, those they deem to be “fraudulent,” or those under thirty seconds—without a provision affirming their right to do so in their licensing deals.

The administrative copyright rulemaking board ultimately adopted Apple’s proposal excluding streams under thirty seconds in full and adopted a modified version of Apple’s proposal to exclude fraudulent streams (those “not . . . initiated or requested by a human user”) from royalty-bearing streams.¹⁷³ This is fascinating, as previously, the § 115 compulsory license recognized no such carve-out for streams under thirty seconds, and it did not give a platform the ability to use algorithms to determine what constitutes a proper stream initiated by a “human user.”¹⁷⁴ That the regulations now do both¹⁷⁵ is a direct example of how provisions from private agreements are being translated directly into public law.

And in fact, as the following Part will discuss, public copyright law is changing to either require, or else defer to, the new private copyright.

III. Enshrining the New Private Copyright in the Public Laws

The previous Part detailed why and how digital platforms like YouTube and Facebook began shifting away from relying on statutory copyright law—the DMCA safe harbor—in favor of a series of confidential licensing agreements that rewrote statutory and common law copyright. More worryingly, these agreements, by nature of their confidentiality clauses, are obscured from public oversight. But while such license agreements were once voluntary, and perhaps even extraordinary, changes to copyright laws could make such private contracting a requirement, enshrining the new private copyright in the public law.

This Part begins by reviewing Europe’s recently adopted Article 17, exploring how and why a European law that mandates private contracting has massive ripple effects for user-generated expression in the United States. It will then discuss two recent changes or lobbied-for changes in U.S. copyright law: one that would remove licensing negotiations between large music publishers and technology platforms from antitrust oversight by the Department of Justice, which had been overseeing such licensing activity since the 1940s, and another that removed all considerations of public policy or the public interest in setting the compulsory licensing rates paid by digital streaming companies. Both of these changes or contemplated changes share, with Europe’s Article 17, the same contours of the new privatization: a press release–ready tale of remuneration for artists that obscures the deeper implications of replacing public copyright law with privately made agreements.

A. Article 17 and Mandated Private Licensing

At the beginning of the new millennium, Europe had also adopted its own, similar version of the DMCA’s § 512.¹⁷⁶ Like § 512 of the DMCA, Article 14 of the Directive on Electronic Commerce provides that service providers are “not liable for the information stored at the request of” users, so long as the platform lacks knowledge of infringing activity and it removes infringing content when notified.¹⁷⁷

And just as copyright holders dislike § 512, they, too, have long lobbied for changes to Article 14 in Europe.¹⁷⁸ In 2015, the push to make a change to a law viewed as unduly friendly to technology behemoths gained greater traction. In a press release issued that December, the European Commission (EC) set forth an action plan to modernize European Union (EU) copyright rules, citing a “growing concern about whether the current EU copyright rules make sure that the value generated by some of the new forms of online content distribution is fairly shared, especially where right holders cannot set licensing terms and negotiate on a fair basis with potential users.”¹⁷⁹

Finally, in early 2019, the European Parliament adopted Article 17 of the Digital Single Market Directive (DSM), which replaced the existing regime of safe harbor protection with a direct licensing obligation.¹⁸⁰ The DSM’s stated purpose centered on licensing: the Directive was clear that it was intended to “facilitate the clearance of rights”¹⁸¹ and to “facilitate certain licensing practices.”¹⁸² By doing so, the DSM reasoned that it was “stimulat[ing] innovation, creativity, investment and production of new content”¹⁸³ and “keeping a high level of protection of copyright.”¹⁸⁴ By requiring certain service providers—namely, only those whose primary business model is to “store and enable users to upload and share a large amount of copyright-protected content,”¹⁸⁵ in essence, the YouTubes and the Facebooks of the world—to get licenses to cover the conduct of their users, the DSM inverted the statutory safe harbor rules that have long governed the internet.¹⁸⁶ In essence, it turned the rising prominence of privately made copyright laws that Part II of this Article described into a statutorily mandated requirement.

While it remains to be seen whether the United States will follow suit in its own § 512 reform—and current hearings on § 512 reform suggest some changes to § 512, perhaps remade in the European image, are possible¹⁸⁷—the fact is, in our new global economy, changes to ex-U.S. law inevitably create ripple effects within the United States as well. The general idea that EU laws invariably become global laws as multinational companies voluntarily extend EU rules to govern their global operations has been termed by Professor Anu Bradford as the “Brussels effect.”¹⁸⁸ Certainly, companies will need to develop effective global content management tools that can accommodate more expansive laws, such as Article 17. The largest companies, such as YouTube and Facebook, already have such measures in place. Facebook’s Rights Manager, for example, has digital fingerprinting and matching technology for images, audio-only, and audiovisual works.¹⁸⁹ Further still, the tool allows individual (small) content owners to upload their works and accompanying metadata to the Rights Manager database, specifying whether rights apply on a worldwide basis or only in certain territories.¹⁹⁰ Giving individual creators the ability to authorize and manage copyright permissions through Rights Manager on a global basis thus cannily solved the concern, voiced by some scholars prior to Article 17’s adoption, that obtaining rights for millions of images would prove to be impossible.¹⁹¹

Though such developments may be heralded as achieving precisely what the architects of Article 17 intended—remuneration, or at least some form of control, for the millions of little creators, authors, and artists who can now benefit from the extra royalties generated by a mandatory licensing system¹⁹²—the biggest beneficiaries are not individual creators, but large corporate content-holders, whose licensing deals are worth hundreds of millions of dollars.¹⁹³ Indeed, as discussed in greater depth infra in Part III.B, when parties choose direct licensing instead of government-regulated licensing, there is little accountability or requirement as to how individual songwriters or artists actually get paid.¹⁹⁴

For example, a major complaint upon the leak of the Facebook music publishing contract for indie artists was that the flat-fee deal made payouts to each music publisher based upon their pro-rata market share, meaning indie music publishers received a total pool of $45 million to divide amongst all independent music publishers.¹⁹⁵ Another pool of money, deemed the “Major Pool,” for the major music publishers—Sony, Warner, and Universal—was comprised of an undisclosed amount.¹⁹⁶ That payments were made on a flat-fee, market-share basis squares perfectly with the fact that Facebook has had a difficult time monetizing content.¹⁹⁷ More importantly, the deals explicitly disclaimed any procedure for just how the music publishers would pay the individual songwriters and musicians whose works were being exploited. Indeed, in a section titled “Downstream Payment,” the agreement states: “Facebook will not be in a position to provide track-level or composition-level reporting during the Term. It shall be Publisher’s responsibility to determine how to allocate, if necessary, any Fee it receives to all downstream royalty participants in respect of Publisher Compositions.”¹⁹⁸

Yet as the following Section discusses, rather than distribute royalties downstream to songwriters, publishers often keep—for a variety of reasons—large amounts of the proceeds from direct deals for themselves. Indeed, that is the very reason they prefer unregulated direct deals rather than more regulated markets.

B. Removing Streaming Rights from Department of Justice Oversight

Musical composition owners enjoy the right to publicly perform their copyrighted works.¹⁹⁹ The right is implicated any time a song is played in a restaurant, a stadium, a hotel lobby, on terrestrial radio, or streamed on a digital service like Spotify. The market for public performance rights is heavily regulated, as the majority of these rights²⁰⁰ are held by two performing rights organizations (PROs), ASCAP and BMI, both of which have been subject to consent decrees since the 1940s.²⁰¹ These decrees are overseen by the Department of Justice’s Antitrust Division and were entered into because of investigations into anticompetitive practices of both PROs.²⁰² Under the decrees, ASCAP and BMI must grant a blanket license to all the works in their repertoires and to any user that applies for one—whether it is a restaurant, a television station, or Spotify.²⁰³ In the event the parties cannot agree on a reasonable royalty for the license, they can seek relief in court before one of two district court judges in the Southern District of New York.²⁰⁴ As an additional restriction imposed by the DOJ, the SDNY judge, in setting rates, was originally barred from considering certain evidence from the “marketplace,” like sound recording royalties (which are negotiated on the free market)—a prohibition that has since been repealed by the Music Modernization Act (in order to, what else? Better approach marketplace rates).²⁰⁵ The combination of these two DOJ consent decree features—the license-on-demand provision and the rate-court provision—was intended to address the DOJ’s concern that ASCAP and BMI were subjecting licensees to a devil’s choice: “[A]ccept a license from the [performing rights organization] upon any terms and conditions imposed by [that organization] or subject themselves to numerous infringement suits.”²⁰⁶

Beginning in 2011, however, three of the United States’ four largest music publishers²⁰⁷ began pursuing a strategy of selective, or partial, withdrawal of “new media” rights from ASCAP and BMI.²⁰⁸ Withdrawing “new media” rights means that digital services like Pandora and Spotify would no longer be able to take advantage of ASCAP and BMI’s blanket licenses-on-demand. Streaming rates instead must be negotiated in the free market, removed from DOJ oversight and the restrictions of the consent decrees.²⁰⁹

The direct negotiations that occurred in the aftermath of the partial withdrawals resulted in dramatic rate increases for digital services, which led the services to challenge the permissibility of withdrawal in court.²¹⁰ The district court, and then the Second Circuit, agreed with the digital services that, under the consent decrees, partial withdrawal was prohibited.²¹¹

If the current consent decrees prohibited partial withdrawal, then perhaps what was needed to better effectuate direct licensing was an overhaul of the consent decrees completely. Thus, in the summer of 2019, the DOJ opened a new review of the consent decrees, in large part to consider whether partial withdrawal for streaming rights should be permitted.²¹² Indeed, partial withdrawal was the banner item on the agenda, as copyright holders such as the National Music Publishers’ Association (the main trade association for music publishers and writers) have focused their reform efforts exclusively on “whether copyright owners should be permitted to ‘selectively withdraw’ digital public performance rights from the repertories of ASCAP and BMI.”²¹³

Then-Assistant Attorney General Makan Delrahim espoused the value of direct licensing over government oversight. “[T]he default,” he stated at a workshop on “Competition in Licensing Music Public Performance Rights,” should be, echoing those magic words copyright holders have long touted, “the free market, not enforcement by government decree.”²¹⁴ Echoing the logic of the laissez-faire economists, Delrahim explicitly invoked the principle that the “market” is where everyone obtains their fair share, their Lockean just deserts, stating that “it is critical that songwriters, composers, and musicians enjoy the fruits of a free market for their creativity.”²¹⁵ Just as in the rhetoric surrounding the passage of Article 17, Delrahim explicitly paints consent decree regulation as oppressive to artists and the promise of partial withdrawal and marketplace transactions freeing for individual creators: “Artists who give us those melodies and songs deserve economic liberty,” Delrahim tells us.²¹⁶

But the incentives-for-authors rhetoric is contradicted by public statements and positions taken by the very songwriters that these changes are alleged to benefit. In reality, direct deals for performance rights negotiated outside the protection of the consent decrees and the PRO licensing system, just as with Article 17’s direct deals for user-generated content, largely benefit corporate music publishers, not individual songwriters. While it is impossible to confirm that the music publishers’ confidential direct deals with streaming services contained a provision similar to the Facebook deal that disclaimed any procedure for or obligation to pay songwriters their share of the enormous value at stake,²¹⁷ this concern was precisely what motivated the Songwriters Guild of America (SGA) to denounce partial withdrawal and direct licensing as “catastrophic” for “songwriters and composers due to obfuscation and oversight inability and failure.”²¹⁸

The SGA made these statements in response to a DOJ call for comments on the continued efficacy of the ASCAP/BMI consent decrees.²¹⁹ In its submission, the United States’ oldest and largest organization for songwriters and composers emphasized that there was “strong disagreement between the songwriter community on the one hand and [the] music publisher members on the other” regarding “the wholly unnecessary extension to music publishers . . . of the authority to engage in the partial withdrawal of rights from the PROs.”²²⁰ Just as this Article has detailed in depth, the SGA was concerned that because direct deals are confidential, they eliminate

any semblance of transparency by music publishers in any direct performing rights licensing deal of their choosing, enabling them to completely obfuscate licensing terms from music creators including such crucial information as the inclusion of advances, administrative fees, equity interests, and other remuneration in which music creators have a rightful expectation to share . . . .²²¹

And just as this Article has done, the SGA has had to rely on leaked information, or tidbits of information disclosed through public litigation, to glean even the slightest hint as to the types of deals that are being struck between large music publishers and large copyright licensees,²²² all of which serve to further enrich corporate interests with no regard for the creative welfare of the actual individual creators whose works are being exploited and whose well-being has long served as the false animating rhetoric for expansive copyright protections.

In one example, the SGA points to a licensing deal struck “outside of the PRO collective licensing system”²²³ between Sony Music and DMX, a popular background music (sometimes referred to as “muzak”—those bland, soothing melodies that play over hotel and conference room speakers)²²⁴ provider.²²⁵ Only through deposition testimony that was made public through subsequent litigation was SGA able to discover that a direct license between DMX and BMI contained an advance of $2.4 million dollars (as well as a subsequent $300,000 payout to BMI).²²⁶ “Were it not for this testimony,” SGA writes, “it is likely that no songwriter or composer (whether or not he or she . . . has works in the Sony music publishing catalog . . . ) would ever have known that Sony had received advances and administrative fees from DMX for the direct licensing of performing rights . . . .”²²⁷

The leaked numbers from the Sony deal are likely “the very tip of the iceberg concerning the economic harm already done to music creators through the direct, opaque licensing of performing rights,” the SGA noted.²²⁸ Direct licensing also presents a number of other ills for individual creators, many of which the SGA cataloged in its DOJ submission.²²⁹ Perhaps one of the most worrying other examples is how direct licensing affects the process by which music publishers recoup the advances they pay to the fledgling songwriters on their roster. In deals struck within the collective PRO licensing system, the PRO will pay the songwriter’s royalty directly to the songwriter.²³⁰ As a consequence, the music publisher is unable to recoup its advance out of these royalties—they belong strictly to the songwriter.²³¹ Direct licenses negotiated outside of the collective PRO licensing system, however, present a new boon to music publishers, as music publishers are free to retain any royalties or other payments paid to them by a direct licensee like Spotify without paying anything to the songwriter.²³²

Indeed, the common refrain that artists do not earn a penny from streaming oftentimes has nothing to do with alleged low royalties paid by streaming services—and everything to do with the fact that, in direct deals generally, royalties are paid directly to music publishers and record labels with no concomitant procedure for accounting to songwriters or musicians.²³³ If a record label or music publisher wishes to use the royalties it receives from Spotify for a stream of an artist’s song to recoup the advance paid to that artist, it is free to do so. As a result, a musician whose song gathers “over 100 million streams on Spotify” but is “not recouped on [their] original contract [with the record label or music publisher] . . . earns nothing from streams of that song.”²³⁴ Of course, record labels and music publishers need not necessarily rely on recoupment to keep most of the fruits of a lucrative licensing deal for themselves. Other leaked agreements, such as one between Spotify and the record label Sony, show that Spotify paid Sony Music up to $42.5 million in advances—and, just like the Facebook agreement, with no provision that provides for an accounting of that large sum to artists.²³⁵ Reporting surrounding the (since removed) agreement confirms that “labels routinely keep [such] advances for themselves.”²³⁶ And these amounts are on top of the equity in Spotify that labels received—which amounted to billions of dollars in value upon Spotify’s IPO.²³⁷ Nothing in the confidential agreements obligated labels to share any of that enormous value with artists—and it was only upon public disclosure of the true worth of those equity shares and subsequent artist complaints that labels were forced to publicly pronounce their commitment to sharing the proceeds with artists.²³⁸

Even as the consent decrees remain in place—for now²³⁹—other regulated areas of copyright are moving towards greater private market deference. The last Section of this Part considers the recently enacted Music Modernization Act and the replacement of the last vestiges of a public interest consideration in the Copyright Act with a market-based standard.

C. The Music Modernization Act: Rejecting Public Interest Considerations in Favor of Private Ordering

The § 115 mechanical license²⁴⁰ is the oldest compulsory license in U.S. copyright law.²⁴¹ It is also one of the few compulsory licenses available under U.S. copyright law.²⁴² Originally enacted to prevent piano roll manufacturers from monopolizing the player piano market, the mechanical license requires owners of musical compositions to license it to anyone who wishes to make a copy.²⁴³ In the digital age, services like Spotify, which make copies of the musical composition in the course of streaming,²⁴⁴ can also take advantage of the mechanical right. That is, a song that is streamed on Spotify is actually two distinct copyrighted works: a sound recording²⁴⁵ and a musical composition.²⁴⁶ Whereas Spotify must engage in individual negotiations with record labels to obtain the streaming rights for the sound recording, Spotify need only follow the procedure set forth in § 115 to obtain a compulsory license to reproduce the musical composition by serving notice upon the copyright holder that it intends to make use of the copyrighted work.²⁴⁷

Rates for the § 115 mechanical license are set by an administrative proceeding before the Copyright Royalty Board (CRB). And for the past four-plus decades—since the establishment of the CRB²⁴⁸—those rates were not based on marketplace rates but rather were determined according to a discrete set of “public-interest” oriented²⁴⁹ policy objectives.²⁵⁰ Those four objectives were:

(A) [t]o maximize the availability of creative works to the public[,]
(B) [t]o afford the copyright owner a fair return for his or her creative work and the copyright user a fair income under existing economic conditions[,]
(C) [t]o reflect the relative roles of the copyright owner and the copyright user in the product made available to the public with respect to relative creative contribution, technological contribution, capital investment, cost, risk, and contribution to the opening of new markets for creative expression and media for their communication[, and]
(D) [t]o minimize any disruptive impact on the structure of the industries involved and on generally prevailing industry practices.²⁵¹

As one CRB judge put it, “[n]otably, section 801(b)(1) does not require the Judges even to attempt to set market rates, or to use market rates to establish ‘reasonable’ rates under the statute.”²⁵² Instead, the four itemized factors each have independent objectives that directly bear on the public interest, ensuring that the public has access to creative works and balancing the respective roles of the copyright holder and the copyright user in making those works available to the public. As noted by the CRB, “[t]hese are not factors necessarily implicated or fully addressed by a market-based analysis.”²⁵³ And as Professor Jacob Victor, who has examined the history of the mechanical license and the role that the § 801(b) policy-based factors play in detail, concludes:

Through this unique approach to rate setting, the regime has fostered technologies that expand and enhance access to existing copyrighted works by allowing these new industries to sometimes take advantage of below-market royalty rates. In this respect, compulsory licensing—like the more frequently discussed copyright limitation, fair use—has provided an essential safety valve for preventing the exclusive rights provided by copyright from overly impeding public access to creative works.²⁵⁴

But in late 2018, Congress passed a new law, the Music Modernization Act (MMA),²⁵⁵ that replaced, in the Copyright Office’s own description of it, “the policy-oriented 801(b)(1) rate-setting standard” with a “new market-based willing buyer/willing seller” standard.²⁵⁶ As the CRB explained, the “two standards are distinguishable by the fact that, unlike [the willing buyer/willing seller standard], section 801(b)(1) does not focus on unregulated marketplace rates.”²⁵⁷

As the D.C. Circuit has previously noted, because the willing buyer/willing seller “standard favor[s] the copyright holders,” rightsholders have long advocated for the abolition of the § 801(b) policy standards in favor of a more rightsholder-friendly willing buyer/willing seller standard.²⁵⁸ The objections coalesced around one touchstone: that § 801(b) seemed to elide the realities and the seeming incontestability of the “market,” a place where every transaction, by nature of having been negotiated by private parties in the “marketplace,” bore some marker of indisputable fairness.²⁵⁹

For example, SoundExchange, the licensing collective that distributes sound recording royalties for satellite and digital radio services, has consistently “urged Congress to establish rate standard parity so [that] all digital services are subject to a ‘willing buyer, willing seller’ or fair market value royalty rate standard[].”²⁶⁰ (In excising the four policy-based objectives of § 801(b) from the statute, the changes created by the MMA benefitted not just music publishers receiving mechanicals for interactive streaming, but also sound recording owners, who were previously subject to § 801(b) rates for certain digital radio and satellite radio transmissions.)²⁶¹ The § 801(b) policy standard, SoundExchange argued, allowed satellite radio services to pay “below-market value royalty rate[s]” for years while “recording artists and rights owners . . . subsidized these companies’ growth.”²⁶² Indeed, SoundExchange had long contested the CRB’s use of the § 801(b) policy factors, including the third factor (which allows the CRB to reduce royalty rates by favoring technological contributions made by services like SiriusXM, which invests substantially in technological infrastructure) and the fourth factor (which allows the CRB take into consideration the disruptive effect of high royalty rates on copyright users).²⁶³ In appeals of the CRB’s § 801(b) rate-setting before the D.C. Circuit, SoundExchange referred to these policy objectives as “trump cards” used to “reduce . . . market-based rate[s].”²⁶⁴

While left unsaid, the copyright industry’s frequent evocations of the importance of honoring a “market-based” rate—and, concomitantly, the injustice of a rate that deviates from it—once again echo the logic of the laissez-faire advocates of the twentieth century, who viewed the markets as a “domain of freedom where . . . all were rewarded ‘in proportion to their just deserts.’”²⁶⁵ Finally, then, the MMA’s replacement of the § 801(b) policy standard with a willing buyer/willing seller rate wrote the last of the public policy-based rates out of copyright law’s compulsory licensing statutes,²⁶⁶ representing the privileging of private ordering, instead.

* * *

Private ordering around and in the shadows of the law has been, and will always continue to be, an inextricable fact of life. Yet as the burgeoning “platform law” literature showcases, private ordering by large digital platforms raises serious questions about power and accountability.²⁶⁷ But even as scholars have theorized a triangular model of speech regulation for the digital age, adding platforms as the third actor, this Article has argued that no account of speech regulation on the internet can be complete without consideration of the powerful copyright owners who control large swaths of speech that users transmit every day. Unlike digital platforms, they do not own the infrastructure of communication—but they do own large parts of the communicatory contents. How they, in concert with platforms, choose to define the boundaries of permissible speech online has, as Part II described, important implications not just for users but also for copyright law. Part IV of this Article concludes by offering some proposals for how users can demand better elucidation of the new rules of platform copyright—and how they can take part in shaping and changing its contours.

IV. Making the Private Public Again

In the late 1990s and early 2000s, content holders—the TV, film, music, and software industries—felt threatened by advances in technology, the rise of the internet, and the seemingly frictionless and costless ease of copying. They began a campaign in both Congress and the courts to enforce shrinkwrap agreements, shut down peer-to-peer technologies, and cement the use of digital rights management tools by creating legal liability for circumventing them.²⁶⁸ It worked, in a sense: Napster, Limewire, and Grokster all fell.²⁶⁹ The DMCA validated the use of private architectures of control over how owners of DVDs and CDs could use those works.²⁷⁰ And despite reasoned arguments against them,²⁷¹ the Seventh Circuit upheld the enforceability of shrinkwrap agreements in a decision that was celebrated by some as validating the importance of free contracting in the face of unprecedented technological change.²⁷² In this, as with all things in the copyright wars,²⁷³ new technology created both a threat and an opportunity—as the argument goes, technology changes things, and the law needs to change to better and more perfectly protect property rights.

Then something odd happened. As it turns out, attempts to press consumers into submission mobilized them into rebellion instead. Copyright users such as purchasers and file-sharers—long thought of as poorly organized and diffuse (as compared to copyright holders, who were concentrated in industries with immense lobbying power)—began to protest such obvious and clumsy²⁷⁴ incursions on how they could view or share content, raising public awareness and forming their own countervailing faction under the umbrella term “Access to Knowledge.”²⁷⁵ College students were suddenly talking about copyright law restrictions “with something like the reverence that earlier generations displayed in talking about social or racial equality.”²⁷⁶ Even as scholars astutely described the penetration of digital rights management tools as a market replacement for copyright’s public law, some expressed doubt as to their ultimate efficacy.²⁷⁷

Thus, throughout the 2010s, public backlash and opposition neutralized the seeming expansion of copyright, creating a sort of stasis. The flurry of public-interest-led lawsuits challenging the prohibition against circumventing digital rights management tools abated.²⁷⁸ The Recording Industry Association of America was no longer suing private citizens for copyright infringement.²⁷⁹ The rise in legal streaming services returned revenues back to a music industry that had sustained prolonged losses in the era of Napster and Limewire and other illegal peer-to-peer networks.²⁸⁰ Even copyright scholarship itself struck a less apocalyptic tone.²⁸¹ Instead of hand-wringing, scholars in favor of users’ rights seemed almost celebratory—all turned out decently, after all.²⁸² And if the mentality of young people growing up in the 1990s and 2000s had been one of “information just wants to be free,”²⁸³ when I presented that same argument to my students in 2017, I was met with confused looks—the battle of ideologies was over. The number of raised hands in response to a question about how many students used bit-torrent or other illegal peer-to-peer services diminished between 2015 and 2016, and then again from 2016 to 2017, and so on. These days, everyone pays for something, or, chances are, multiple things: Spotify, Apple Music, Netflix, Hulu.

The rapid rise and affordability of streaming services had put out the fire, both on the left and on the right. There is the feeling of satisfied complacency in the air, fed by an unprecedented glut of easy-to-access, well-priced content. The advances of the free copyright and Access to Knowledge movements, so prominent across college campuses in the 1990s and at the turn of the millennium,²⁸⁴ have instead given way to a call, across social media and even among some scholars, for more copyright, grounded in what users perceive as unfair freeriding off the intellectual labors of others.²⁸⁵ Words like “appropriation” and “theft” were celebrated in the 2000s because they referred to a rising-up against coercive rightsholders.

The same words are now wielded on social media to denounce what users perceive as unfair freeriding off the intellectual labors of others.²⁸⁶ In this, copyright law itself—as a network of rules, much of it public (common, statutory), but increasingly privatized on the largest internet platforms—has seemed to disappear altogether. If the major goal of this Article has been to shed light on the new private copyright, this final Part puts forth some proposals for reform: First, it recommends including copyright content moderation policies in generalized calls for content moderation reform and transparency. It then looks to developments in Europe that place certain aspects of public copyright law beyond the reach of private contracting, turning default rules into immutable rules. Finally, it suggests that users—nonparties to the very contracts that purport to govern their actions—may reinsert themselves in the process through litigation.

A. Copyright Moderation Decisions Are Content Moderation Decisions

As calls for platforms to make their content moderation decisions more transparent have intensified, Congress has introduced bipartisan legislation that would update § 230 of the Communications Decency Act to “make platforms’ content moderation practices more transparent and hold those companies accountable for content that violates their own policies or is illegal.”²⁸⁷ The proposed PACT Act, for example, would

[r]equir[e] online platforms to explain their content moderation practices in an acceptable use policy that is easily accessible to consumers; [i]mplement[] a biannual reporting requirement for online platforms that includes disaggregated statistics on content that has been removed, demonetized, or deprioritized; and [p]romot[e] open collaboration and sharing of industry best practices and guidelines through a National Institute of Standards and Technology-led voluntary framework.²⁸⁸

It is unclear how, or if, the PACT Act considers the moderation (including through blocking or muting) of works incorporating copyrighted content as included in its ambit. Copyright content moderation practices should either be included in the PACT Act or else similar copyright legislation should be introduced to likewise require platforms to publish clear guidelines for how consumers may use copyrighted content on digital platforms. Such legislation must also make clear that it takes precedence over any applicable confidentiality provisions in private agreements.

In short, any attempts to reform content moderation—a politically salient issue²⁸⁹—must include copyright content moderation within their ambit. Copyright moderation decisions are content moderation decisions.

B. Certain Copyright Default Rules as Immutable Rules

If this Article has detailed how public legislation driven by the interests of concentrated industry groups can end up deferring to private markets by couching itself in the rhetoric of artists’ rights, it does not mean that one should abandon faith outright in the legislative process. One example from the EU of how individual member states have transposed Article 17 of the DSM²⁹⁰ provides a blueprint for how a regime of privatized license agreements may accommodate existing public law by making certain statutory rules, such as fair use, immutable—rules that cannot be contracted around.²⁹¹

Article 17 directs individual member states to “ensure that users . . . are able to rely on” certain existing exceptions to infringement in the law: use of copyrighted content for purposes of “quotation, criticism, [and] review” or for “caricature, parody or pastiche.”²⁹² Much of the public criticism surrounding this provision, which was meant to ensure that the license agreements required by the Directive did not contravene important free speech principles,²⁹³ focused on how individual member states would transpose the provision in a workable manner.²⁹⁴ Most notably, critics worried that the provision lacked teeth—platforms would have little incentive to underblock rather than overblock.²⁹⁵

Germany’s implementation of the law was the first to contain a series of improvements largely welcomed as safeguarding user rights.²⁹⁶ Most notable among these is a provision which actually provides a means of collective redress against platform overblocking by permitting associations representing user rights²⁹⁷ to bring a legal claim against a service provider for repeated and wrongful blocks of authorized uses.²⁹⁸ Likewise, Germany’s implementation carefully sets out a list of presumably authorized uses: statutory rules on permitted expression that cannot be trumped by private contracting.²⁹⁹

Any subsequent U.S. reform of § 512 (including any contemplated use of filtering technology) must likewise provide legal standing for users to police against arbitrary and overexpansive blocks. More importantly, like the German implementation of Article 17, the statute must make clear that private filtering technology and any license agreements that those filters operate under cannot trump certain existing rights in the Copyright Act, such as fair use. Instead, any blocks of user-generated content must be carried out in accordance with statutory law. In this manner, certain well-established copyright rules—such as the fact that parodies utilizing copyrighted works are permissible even if a copyright holder finds the parody offensive or in poor taste—become immutable rules, rules that cannot be contracted around. Thus, contractual terms that attempt to route around those rules would be void and unenforceable.

C. Reinserting the User

This Part began by considering the notable anti-copyright countermovement of the late 1990s and early 2000s. Individual users coalesced around flashpoints as obvious as peer-to-peer filesharing and as seemingly benign as coffee machines. In the latter case, the coffee company Keurig had enlisted an unlikely ally—copyright and patent law—to prevent customers from using non-Keurig-branded coffee products in its machines.³⁰⁰ In the mass outcry and litany of public interest lawsuits that ensued, much of the litigation strategy focused on bringing class action claims on behalf of ordinary consumers under state consumer protection statutes.³⁰¹ These state statutes are often referred to as “mini attorneys general”³⁰² statutes precisely because they empower private citizens to bring claims on account of the public interest.³⁰³

If consumer expectations are unsettled—when Facebook livestreams are muted,³⁰⁴ when streams of songs are improperly tracked,³⁰⁵ when legitimate criticism or discourse about culture is silenced³⁰⁶—consumers need not fumble helplessly amongst online forums³⁰⁷ for answers and redress. Instead, litigation itself can act as a transparency tool, shedding light on the confidential licensing terms that circumscribe user speech.³⁰⁸ Such litigation can act in tandem with, or even be superior to, the legislative solutions posed above that deem certain statutory copyright provisions immutable rules. Indeed, a simmering strand of common law copyright doctrine—broadly known as “copyright misuse”—suggests that because copyright law has important public policy considerations, private attempts to contract around it may be struck down as void against public policy.³⁰⁹ Consider a user whose video was “demonetized” due to a violation of YouTube’s copyright terms of service. Typically, “creators” on YouTube—famous influencers with large followings—are able to receive a portion of YouTube’s advertising revenue from views of that video.³¹⁰ YouTube, however, may “demonetize” creators who violate the platform’s terms of service, including its copyright policy.³¹¹ In cases where another copyright holder has claimed that it owns the copyright in a portion of the creator’s work, the creator’s video may be “demonetized” while the other copyright holder gets to share in monetization proceeds.³¹² The system is prone to error (and potential wrongful claims), such as when a famous YouTube creator tweeted screenshots that her videos were demonetized because someone else claimed copyright protection in the number “50.”³¹³ To avoid losing out on sharing in ad revenue, YouTube creators may “avoid making fair use of copyrighted material they want to use in their work, and endlessly edit and re-edit lawful expression.”³¹⁴

All demonetized creators, then, might bring a class action lawsuit, arguing that YouTube’s prohibition on monetizing even fair uses of copyrighted content contravenes public policy, rendering statements such as that YouTube “support[s] the free flow of ideas and creativity” misleading advertising under state consumer deception laws.³¹⁵ Similar suits have gained traction in recent months: for example, a nonprofit advocacy group recently brought suit against Facebook, alleging that the company violated state consumer protection law when it claimed in testimony to Congress that it removes any harmful content immediately, despite evidence that large swaths of hate speech remain available on the site.³¹⁶ As users everywhere become frustrated by the opaque content moderation policies of the largest internet platforms, they may increasingly turn to individual state mini-attorney-general statutes to force change.

After all, if the legislative process has shown itself to be suspect through industry capture,³¹⁷ then pursuing redress through the judicial system, by aggregating the thousands of small claims of everyday consumers, is one way to counterbalance powerful and concentrated lobbyists. Indeed, if one vision of the class action mechanism sees it as a regulatory device—to deter misconduct and enact wide-ranging social change—then class actions can in effect act as “a regulatory process [that] furthers the design and administration of public policy.”³¹⁸ And even if such suits face an uphill battle to a full adjudication on the merits, the mere filing of a class action complaint itself may serve as a threat and deterrent to overbroad contractual rights that tread on user expression—aggregated user claims, after all, mean increased exposure.³¹⁹

Conclusion

The mushrooming literature on the law of the platform—examining how large digital intermediaries create their own substantive rules and procedures that, by nature of the sheer amount of content and users they apply to, become governmental and legal in nature—has been, for the most part, markedly pessimistic about the “quandary of ensuring democratic accountability and legitimacy when private organizations serve as regulators.”³²⁰ On the other hand, copyright scholarship has largely focused on the internet’s democratizing effect on copyright industries formerly dominated by a select few record labels, movie studios, and publishers. In this sense, digital technologies have led to more content, diffuse and varied individual creators, and new distribution platforms for the amateur creator.³²¹ The rise of new technologies made it easier and cheaper for everyday individuals to create content, the classic story goes.³²² In turn, barriers to distribution have also disappeared, as a number of platforms—YouTube, Facebook, TikTok, and self-publishing—allow creators to bypass traditional gatekeepers such as record labels, movie studios, or book publishers.³²³ The end result is more varied types of, and cheaper, content than ever.³²⁴ We are living in the golden age of creativity.³²⁵

But increasingly, today’s digital fora look like the new gatekeepers. While this Article has highlighted how a consideration of the so-called “platform law”³²⁶ is incomplete without an accurate understanding of how platforms moderate copyright disputes, this Article has also argued that understanding copyright content moderation demands a different framework than traditional content moderation frameworks. Whereas Professor Jack Balkin has theorized a triadic model of speech regulation—speakers, governments, and digital intermediaries³²⁷—this Article has proposed a fourth player: copyright owners who do not control the infrastructure of the communication but claim property rights in the communicatory contents.

As this Article has detailed, the recent rise in private contracting between large content holders and concentrated technological intermediaries like Facebook and Google has supplanted substantive, public copyright law with a series of privately made rules that are obscured from public view. Little is known, and far less has been written, on just what those rules are. Using both public filings and confidential leaked documents, this Article attempted to piece together some clues as to what the new privatized copyright looks like—how it replaces and contravenes substantive public law, and how it may come to influence and shape substantive public law, as well.

Internet law scholars have previously pointed out that governments “are leveraging the infrastructure of private ordering . . . in order to carry out their own policy preferences.”³²⁸ Copyright holders, too, are keen on the benefits of exercising control through a technological intermediary—be it a social media network or a streaming service. Indeed, as Part III of this Article discussed, they are working on rewriting the public copyright law to make such private contracting a requirement or, in other cases, to shield public copyright rulemaking from public policy considerations. Creative speech on the internet now finds itself at a curious precipice: a seeming glut of low-cost or free content, much of which is created by, and distributed to, users—yet which is increasingly regulated by an opaque network of rules created by a select few private parties.

These recent developments threaten the democratizing potential of the internet in a way very different from how most copyright scholars have conceived of internet-facilitated democratization (that is, the idea that anyone can now be a creator).³²⁹ Rather, the developments endanger the very concept of democratization as collective governance. If the rules of content creation on the internet come increasingly to be governed by invisible contracts negotiated between a select handful of large corporate conglomerates, then the meaning of democratization as the anything-goes freedom of content creation on the internet may be threatened. This Article concluded with some proposals for how the unruly masses that the new private copyright governs can reinsert themselves in the copyright process—by first pushing for a greater understanding of what the new privately made rules are before challenging their legality and reaffirming the primacy of public law in the courts. If they don’t—if we don’t—the democratized (in all senses of the word) digital renaissance may end up a failed experiment.

* Assistant Professor, UCLA School of Law. Thanks to BJ Ard, Barton Beebe, Brett Frischmann, Jeanne Fromer, Kristelia García, Eric Goldman, Orin Kerr, Mark Lemley, Doug Lichtman, Glynn Lunney, Mark McKenna, Peter Menell, Neil Netanel, Ruth Okediji, Kal Raustiala, Jennifer Rothman, Guy Rub, Pamela Samuelson, Matthew Sag, Jessica Silbey, Seana Shiffrin, Christopher Sprigman, Rebecca Tushnet, Jacob Victor, Christopher Buccafusco, Sonia Katyal, Jessica Litman, Robert Merges, Tejas Narechania, Felix Wu, and the participants at the 2022 Harvard/Stanford/Yale Junior Faculty Forum, the 2022 AALS Emerging Voices in Intellectual Property Panel, the 2021 Southeastern Association of Law Schools Annual Conference, the 2021 Works-in-Progress Conference, the 2021 Intellectual Property Scholars Conference, the 2020 Copyright Scholarship Roundtable, the 2020 Junior Intellectual Property Scholars Association Workshop, the Arizona State University Sandra Day O’Connor School of Law Faculty Colloquium, the Cardozo Law School Intellectual Property Colloquium, the Berkeley/UCLA Junior Faculty Exchange, and the UCLA Law Junior Scholars Workshop for generous comments and dialogues on earlier drafts. Elizabeth Anastasi and Tyler Emeney provided superb research assistance. The author previously served as Lead Counsel for Music Licensing at Facebook, Inc. (now Meta), and represented Spotify USA Inc. while an attorney at Mayer Brown LLP. All information herein is from the public record. Finally, my deep thanks to William Ellis, Taylor Wilson, and the editors at the Michigan Law Review for their insightful commentary and hard work in preparing this Article for print.