I. Introduction

The law of secondary copyright liability and the law of secondary patent liability largely parallel each other. Each has similar forms of liability, such as contributory infringement and inducement, and each has at least similarly worded tests for those forms of liability. On the other hand, a secondary copyright case will generally look quite different from a secondary patent case. Most secondary patent infringers act in a way that targets a particular patent or group of related patents. Secondary copyright infringement, though, often arises in the context of technologies or services that work across all copyrighted works. Secondary copyright liability raises issues of platform liability in ways that secondary patent liability usually does not.

The difference between secondary copyright and secondary patent cases is important because standards designed to apply to non-platform cases may poorly suit platform cases. Non-platform secondary liability cases often hinge on questions of knowledge or fault, while incentives and efficiency play a greater role in platform cases. The result is that the current rules of secondary copyright liability are framed too much in terms of mens rea and fault and not enough in terms of avoiding both under- and over-deterrence.

Adopting a more platform-oriented perspective to secondary copyright liability might counsel toward shifting the law in such cases in terms of both substance and remedies. On substance, we could limit secondary copyright liability to exact copies, but then require some level of reasonableness rather than merely the removal of known infringing works. On remedies, we could eliminate statutory damages for secondary infringement, and instead rely on disgorgement of profits directly attributable to infringement, for example in the form of ad revenue. Schemes like these have the potential to achieve a better balance than the current law does between harnessing the power of intermediaries to police infringement and avoiding a chilling effect on noninfringing activities.

Shifting the law of secondary copyright liability, however, is not the central argument of this Article. There are competing goals in any secondary liability scheme, and adjustments inevitably involve trading off one goal for another: a change could discourage infringement while also blocking some noninfringing uses, or vice versa. Perhaps the status quo, with its mix of common law origins, statutory safe harbors, and private ordering, does a reasonably good job of striking an appropriate balance. Perhaps courts have done a reasonably good job at reshaping old terms and old tests to fit the platform economy. Even then, it is worth understanding why the status quo works (if it works), when to encourage or resist judicial reinterpretation of applicable tests, and how to design and evaluate proposals for legislative change. For doing any of these things, viewing secondary copyright liability through the lens of platform liability principles is crucial.

In what follows, Part II explains the law of secondary liability in patent and copyright law. Part III describes how secondary copyright cases are platform cases, why law developed in non-platform contexts may be a poor fit for platform cases, and which liability principles gain heightened importance in the platform context. Part IV explores some possible implications for designing secondary copyright liability and for remedies, and it concludes with some thoughts about why a proper understanding of the structure of secondary copyright law matters.

II. Secondary Liability in Patent and Copyright Law

At least at a formal level, secondary liability is structurally similar in patent and copyright law.

A. Patent Law

Secondary patent liability is codified in the patent statute. 35 U.S.C. § 271(a) describes the circumstances for direct infringement.[1] Section 271(b) then states a cause of action for inducing infringement and § 271(c) for contributory infringement, as follows:

(b) Whoever actively induces infringement of a patent shall be liable as an infringer.

(c) Whoever offers to sell or sells within the United States or imports into the United States a component of a patented machine, manufacture, combination or composition, or a material or apparatus for use in practicing a patented process, constituting a material part of the invention, knowing the same to be especially made or especially adapted for use in an infringement of such patent, and not a staple article or commodity of commerce suitable for substantial noninfringing use, shall be liable as a contributory infringer.[2]

A key issue in interpreting the secondary patent liability provisions has been what level of knowledge or fault is required to trigger those provisions.[3] Under § 271(c), for example, the statute is not entirely clear on what a defendant needs to “know” in order to know that something is “especially made or especially adapted for use in an infringement of [a] patent.”[4] On the one hand, it could be enough to know that the component is “especially made or especially adapted” for a particular use that then turns out to infringe. On the other hand, perhaps the defendant also needs to know that the use for which the component is especially made or adapted is in fact infringing. Early on, the Supreme Court adopted the latter, stricter view of knowledge, holding that “§ 271(c) does require a showing that the alleged contributory infringer knew that the combination for which his component was especially designed was both patented and infringing.”[5]

Similarly, § 271(b) does not specify what it means to “actively induce” infringement. Again, one possibility is that it is enough to induce conduct that turns out to infringe, while another is that the defendant must know that the induced conduct is infringing. And again, the Court adopted the latter view, relying on the Aro case.[6]

Secondary copyright liability is not expressly codified, but its judge-made structure parallels that of patent law in terms of both the forms of liability and the focus on knowledge or intent. “Traditionally, ‘one who, with knowledge of the infringing activity, induces, causes or materially contributes to the infringing conduct of another, may be held liable as a “contributory” infringer.’”[7] The knowledge in question must be that of “specific infringing material,” rather than a system’s general capacity for infringement.[8] Moreover, by analogy to the “staple article of commerce” carveout to patent contributory liability, “the sale of copying equipment, like the sale of other articles of commerce, does not constitute [copyright] contributory infringement if the product is . . . capable of substantial noninfringing uses.”[9]

Alternatively (or as a species of contributory liability), copyright law also recognizes secondary liability for inducement. “[O]ne who distributes a device with the object of promoting its use to infringe copyright, as shown by clear expression or other affirmative steps taken to foster infringement, is liable for the resulting acts of infringement by third parties.”[10] Such a rule is based on a finding of “culpable intent” and is a form of “fault-based liability derived from the common law.”[11]

While secondary copyright liability itself is not codified, the Digital Millennium Copyright Act (DMCA) safe harbors codified at 17 U.S.C. § 512 are also a key part of the secondary liability landscape. Here too, the safe harbors, especially the often invoked § 512(c), draw heavily from the preexisting common law rules on contributory liability and their focus on the defendant’s knowledge. Section 512(c) provides that the safe harbor applies only if the service provider:

(A)(i) does not have actual knowledge that the material or an activity using the material on the system or network is infringing;

(ii) in the absence of such actual knowledge, is not aware of facts or circumstances from which infringing activity is apparent; or

(iii) upon obtaining such knowledge or awareness, acts expeditiously to remove, or disable access to, the material.[12]

In some instances, the parallel between secondary patent and secondary copyright liability has been the result of explicit borrowing from the patent statute. The Sony Court drew from “the patent law cases to which it is appropriate to refer because of the historic kinship between patent law and copyright law.”[13] The Grokster Court explained that “[f]or the same reasons that Sony took the staple-article doctrine of patent law as a model for its copyright safe-harbor rule, the [patent] inducement rule, too, is a sensible one for copyright.”[14] Even without explicit borrowing, the commonalities between secondary patent and secondary copyright liability are unsurprising, given that each flowed from earlier common law tort principles.[15]

To be sure, the parallel here should not be overstated. Vicarious liability is formulated somewhat differently in patent and copyright law and is not based on a finding of knowledge or intent.[16] True vicarious liability cases seem to be somewhat uncommon in patent law. Instead, courts have drawn on “general principles of vicarious liability” to “determine if a single entity directs or controls the acts of another” in a context in which more than one actor is involved in performing the steps of a claimed method—this is the issue of “divided infringement.”[17] In copyright law, vicarious liability has been extended beyond its common law origins in agency relationships to encompass “cases in which a defendant ‘has the right and ability to supervise the infringing activity and also has a direct financial interest in such activities.’”[18] The same standard (or at least the same words) are used as part of the requirements for invoking the § 512(c) safe harbor.[19]

III. Platform Cases and Platform Liability

Secondary patent and secondary copyright law are quite similar to each other, and yet secondary patent and secondary copyright cases are not. Many secondary copyright cases are platform cases. The distinction matters because law designed around non-platform cases may not be oriented around the right principles for platform cases.

In the paradigmatic secondary patent case, some specific act of infringement has occurred, and the question is whether the potential secondary infringer had enough involvement in or fault as to that particular act of infringement so as to be held responsible for it. If such a defendant is liable at all, they will be liable for infringing a particular patent or related set of patents relevant to the particular infringing act at issue. Aro, for example, involved a manufacturer of fabric components for convertible automobile tops and a patent on convertible tops.[20] If such a manufacturer is a secondary infringer, it would infringe that particular patent, or perhaps some other patent covering the same technology.

In many secondary copyright cases, by contrast, the potential secondary infringer is providing a product or service that works across all copyrighted works, or at least all works in a particular medium. YouTube works across all videos. Peer-to-peer file sharing networks are agnostic about the files being shared—they could be any song, video, book, software, or other file. This feature of secondary copyright cases is not specific to the digital world. Dance halls could play any kind of music. Flea markets could sell any kind of physical media. Secondary copyright cases are generally platform cases. They involve an entity that provides to third parties the ability to communicate copyrighted works of those third parties’ own choosing.

The distinction between platform and non-platform cases matters because it changes the appropriate unit of analysis. In non-platform patent cases, the focus is on a specific instance of infringement and on assessing whether the defendant bears enough responsibility for that infringement to be held jointly liable for it. The model is that of joint tortfeasors in tort law.[21] Such an entity is not being asked to distinguish between infringing and noninfringing uses of its product.

In platform copyright cases, the focus is instead on the platform as a whole. It is not especially meaningful to look at a particular infringement on the platform and ask what the platform’s role in that infringement is. The answer will be largely the same across all of the infringing activity on the platform. Instead, here we do want the platform to distinguish between infringing and noninfringing works, and the key question is how much of an effort to distinguish should be required.

Of course, not every secondary copyright case is a platform case. There are cases involving the liability of investors, for example, in which the secondary copyright defendant is not acting as a platform and may target its activities at specific copyrighted works.[22] The rise of generative AI is potentially creating a new class of cases that, while technologically enabled, are not based on the defendant acting as a platform.[23] In these non-platform secondary copyright cases, traditional liability principles that parallel those in patent law may be appropriate and sufficient. Platform cases, however, call for something different.

B. The Misfit Between Platform Cases and Non-Platform Law

Applying law developed in a non-platform context to platforms can be problematic. Indeed, the platform nature of secondary copyright cases helps to explain why the Supreme Court’s two instances of explicit borrowing from secondary patent liability do not make much sense. Start with Sony and the “staple article of commerce” doctrine.[24] In patent law, the opposite of a “staple article of commerce” is a component “especially made or especially adapted for use in an infringement of such patent.”[25] Note the reference to “such patent,” which suggests that the question is whether the component is tailored for a particular patent, as the convertible top fabrics were in Aro.

In the copyright context, no platform is “especially made or especially adapted for use in an infringement” of a particular copyright. As noted above, platforms function across a wide range of copyrighted works and are not tailored to one work or a small set of related works. Said differently, platforms are always “capable of substantial noninfringing uses,” namely uses involving works in the public domain or uses where the owner of the work does not object. With a component tailored for a particular patented context, if it is not used in that context, it is probably good for nothing other than being a very fancy doorstop (which would not be a “substantial noninfringing use”). A platform is always capable of being more than a doorstop.

Thus, if the staple article of commerce doctrine were to be applied in copyright law the same as in patent law, it would function to eliminate the possibility of contributory infringement by platforms. That is, of course, not what has happened. Even in Sony itself, the Court ultimately relied not only on the possibility of authorized recordings, but also on finding that unauthorized time-shifting is fair use.[26] One might think that using a video recorder to make authorized recordings would be a clear substantial noninfringing use that the recorder is “capable of.” If this is not necessarily enough, it’s unclear why not or what more is required. In the Napster case, the Ninth Circuit suggested that the staple article of commerce doctrine merely restricts what counts as constructive knowledge of infringement and is irrelevant when the defendant has actual knowledge of infringement.[27] These formulations of the staple article of commerce doctrine render it either highly indeterminate or largely irrelevant as applied to platforms. In any event, neither formulation bears much resemblance to how the doctrine is applied in patent law.

Turning next to Grokster and inducement: in the patent context, a defendant liable for inducement has generally targeted a particular patent or technology, such that to avoid liability, the defendant must stop its activity altogether (or get a license for it). In Global-Tech, for example, the defendant foreign manufacturer was found to have induced the sale of the patented item in the United States by selling it to the U.S. seller knowing that the item was bound for the United States.[28] To avoid this liability, either the ultimate U.S. sale would need to be licensed or the foreign manufacturer would need to manufacture something else entirely.

In the copyright context, a platform may have induced the infringement of some works, but it is hard to see how it could have induced all of the infringements occurring on the platform, unless “inducement” means something different here than in the patent context. If the patent standard for inducement applied, a platform liable for inducing copyright infringement could be liable for past misdeeds but could continue to operate the platform with a mix of noninfringing works and uninduced infringing works.

But again, that is not how the Supreme Court interpreted the inducement standard when it imported it into the copyright context. Instead, it held that “one who distributes a device with the object of promoting its use to infringe copyright” is liable.[29] According to this standard, there is no need to tie the inducement to particular acts of infringement. Inducing infringement generally, rather than inducing particular infringements, is enough. This comes close to transforming inducement liability from a form of secondary liability for someone else’s act of infringement into itself an independent wrong. The Court recognized that because inducement is at least formally a kind of secondary liability, stating that “the inducement theory of course requires evidence of actual infringement by recipients of the device, the software in this case.”[30] But it was enough that “there [was] evidence of infringement on a gigantic scale.”[31] The Court saw no need to draw any connection between the evidence of unlawful intent and any subsequent infringements.[32] This is perhaps the only sensible way to apply a theory of inducement to platforms. It is not how inducement has been applied in the patent context.

Thus, even where the tests for secondary liability are formally the same, because tests designed for non-platform patent cases don’t work equally well in platform copyright cases, courts have reinterpreted those tests to better fit the platform context.

Indeed, the basic test for contributory infringement is itself an example of a standard that is facially similar between the two regimes but has been interpreted in very different ways. In both patent and copyright law, contributory infringement hinges on a finding of “knowledge.”[33] But while patent law requires a stringent form of actual knowledge for a defendant to be held contributorily liable,[34] in copyright law, courts have found that constructive knowledge is enough.[35] This perhaps reflects at least an implicit recognition of the diminished importance that knowledge plays in appropriately addressing secondary copyright cases.

Something similar has happened within trademark law, where secondary liability principles originally developed for non-platform settings have morphed as courts have applied them in platform settings. Secondary trademark liability is entirely non-statutory, and the vast majority of modern secondary trademark cases follow the Supreme Court’s articulation of the standard for secondary liability in Inwood Laboratories, Inc. v. Ives Laboratories, Inc.[36] The question in that case was whether manufacturers of generic medications that were visually similar to their branded equivalents could be held responsible when pharmacists sold generic pills in bottles labeled with the brand name.[37] The Court held that the applicable standard is that “if a manufacturer or distributor intentionally induces another to infringe a trademark, or if it continues to supply its product to one whom it knows or has reason to know is engaging in trademark infringement, the manufacturer or distributor is contributorially [sic] responsible for any harm done as a result of the deceit.”[38]

Inwood was a decidedly non-platform case, very much in the model of most patent cases. The activities of the generic drug manufacturers were clearly aimed at the trademarks of the brand name manufacturers, and those would be the marks they infringed, if they infringed any at all. The question was how complicit the generic drug companies were in the pharmacists’ acts of infringement. The resulting standard then, unsurprisingly, hinged on questions of knowledge and intent.[39]

The Inwood standard proved to be a poor fit when courts applied it, more recently, to platform cases.[40] In Tiffany (NJ) Inc. v. eBay, Inc., the Second Circuit interpreted the Inwood standard strictly, holding that “general knowledge or reason to know that [a] service is being used to sell counterfeit goods” is not enough for secondary liability and that instead, “[s]ome contemporary knowledge of which particular listings are infringing or will infringe in the future is necessary.”[41] This seems to be the standard that the Supreme Court had in mind in Inwood, or at least what the Court later decided it had in mind.[42] As applied to platforms, however, it led to the complaint that this formulation of the Inwood standard would allow platforms to sit back and do nothing even if there were easy steps they could take to prevent infringement.[43]

Courts have been receptive to that pushback, and since Tiffany, courts have broadened what it means to have sufficient knowledge of infringement, even while claiming not to contradict the Second Circuit’s decision. In one case, the Eleventh Circuit found the owners of a “discount mall” contributorily liable based on having been “put on notice by Luxottica that unnamed subtenants’ [sic] may have been selling counterfeit Luxottica products,” together with the fact that “a cursory visual inspection of 130 booths to see if they displayed what appeared to be counterfeit Luxottica eyewear was not so burdensome as to relieve the defendants of the responsibility to investigate after being [so] informed.”[44] The Eleventh Circuit distinguished the Tiffany case as involving “a marketplace of 100 million listings and eBay’s inability to inspect goods in person and lack of expertise to distinguish Tiffany from non-Tiffany jewelry.”[45]

In other words, the court moved from strict notions of “knowledge” of the sort one might see in secondary patent cases to broader concepts of “constructive knowledge” rooted not so much in considerations of the defendant’s mental state, but rather in efficiency considerations. The mini mall had sufficient “knowledge” because they could easily check their vendors. eBay did not because it would be impossible for them to check their sellers. What either party literally “knew” or didn’t know was not what determined the outcome. Courts have felt constrained to use the Supreme Court’s formulation of the standard from a non-platform case, but they have reinterpreted the terms in that standard to better fit principles of platform liability.

C. Platform Liability Principles

Once we recognize the platform nature of secondary copyright cases, we can use that insight to think differently about the proper design of secondary copyright liability, breaking free of the non-platform patent model focused on knowledge and intent and drawing instead from platform liability principles.

One such principle is that platforms have a propensity to over-block content, particularly when speech is at stake, if platforms and users of those platforms are treated alike.[46] The idea is that a liability regime that has been optimized for original speakers will generally lead to over-blocking if applied identically to platforms. “Optimized” in this context means that the rules strike an appropriate balance between deterring unlawful speech and not chilling lawful speech.[47] When those same rules are applied to platforms, the platforms will generally err on the side of blocking more, rather than less, both because they will usually have less of a commitment to any individual piece of content than the user does and because they may have more difficulty evaluating the lawfulness of the speech than the user has.[48]

In the face of concerns about over-blocking, knowledge or notice may no longer be an appropriate trigger for liability.[49] In the context of 47 U.S.C. § 230, which immunizes Internet intermediaries from liability for content provided by others,[50] courts have uniformly held that the protections of § 230 apply even if the intermediary has knowledge of or has received a notice about the unlawful content.[51] The concern is that otherwise fear of liability would cause intermediaries to block any complained-about content, whether lawful or not, thereby leading to a kind of “chilling effect.”[52]

A different, and somewhat opposing, platform principle is that platforms can be the least cost avoider with respect to policing infringement and/or that they can minimize the transaction costs of licensing content.[53] Platforms are often a bottleneck, such that they can aggregate efforts across multiple instances of infringement, whether of the same or different works, rather than needing to deal with each instance of infringement separately.[54] Platforms can also aggregate licensing transactions across works and across users in a way that is much more efficient than separately licensing each user’s use of each work.[55]

Again, the concept of being a least cost avoider or minimizing transaction costs does not hinge on a platform’s knowledge or intent. It may be cheaper and easier for the platform to develop or acquire the relevant knowledge of infringement than to expect plaintiffs to obtain that knowledge and then pass it along to the platform.

While fixating on knowledge at the expense of a cost-benefit analysis might lead to under-deterrence, it can just as easily lead to over-deterrence, as shown by the cases against Cox Communications.[56] As an Internet service provider, Cox’s options for dealing with infringements on their network are relatively limited. Besides cajoling their subscribers, the main action they can take is to suspend or terminate their subscribers’ internet access entirely.[57] Such an action clearly results in a form of over-blocking, to the extent that all of a subscriber’s activities, infringing and noninfringing, are effectively blocked. Moreover, Cox does not seem like a least-cost avoider here, at least in the sense that, by cutting someone off from the Internet, terminating a subscriber is socially costly.

And yet, in upholding jury verdicts against Cox, the district court found no room in the existing contributory infringement standard for such considerations.[58] The court found that the “provision of high-speed internet service” could constitute a material contribution.[59] And it found that repeated notices gave Cox sufficient knowledge of infringement.[60] In the BMG case, the Fourth Circuit ultimately reversed and remanded for a new trial, finding the jury instructions potentially too permissive as to the level of knowledge required.[61] Cox also made various arguments about why the notices were insufficient.[62] But because these arguments and findings were all tied to the issue of knowledge, none of the back and forth among the courts or parties provided any way to account for the burden to subscribers of having their Internet access terminated.[63] Focusing on knowledge distracted from what was really at stake, whereas platform principles would have taken those interests into account.

Both platform principles described above are premised on not treating a platform and its users as interchangeable. The over-blocking problem arises because platforms and users have different incentives to censor or self-censor content. The least cost avoider analysis relies on there being different costs to enforce IP rights through platforms versus users. It is perhaps unsurprising then that these principles are not consistent with traditional secondary liability rules given that those rules are based on treating direct and secondary infringers as joint tortfeasors that are jointly and severally liable for the resulting infringement.[64] Knowledge or fault plays a key role in traditional secondary liability rules because establishing sufficient knowledge or fault is important when deciding whether someone is a joint tortfeasor.

With respect to platforms, however, it does not make sense to treat platforms as joint tortfeasors with their infringing users. The platforms and users are not acting in concert in the way that the joint tortfeasor model envisions. Other models such as premises liability,[65] products liability,[66] or negligent entrustment[67] are more apt. Each of those models requires an entity to take some level of care in the face of potential misconduct by some other party. A business owner may need to take steps to prevent criminal activity on the business’ premises, for example.[68] Such an entity will face no liability if the misconduct never occurs, even if it fails to take due care, because no harm will have occurred. In that sense, any liability imposed on that entity is derivative of some other wrong. But if the misconduct and harm do occur, liability is based on the entity’s own acts or omissions rather than on the entity taking responsibility for the other party’s misconduct.[69] The business owner and a third-party criminal on the premises are not acting in concert with each other or as joint tortfeasors. Framing platform liability as an independent wrong rather than a component of the user’s wrong more accurately reflects the relative roles of platforms and platform users. So framed, it becomes clear that whether a platform’s acts are independently wrong does not depend solely, or even primarily, on the platform’s knowledge or intent with respect to its users’ wrongs.

IV. Implications

Understanding that secondary copyright cases are often platform cases has potential doctrinal implications. Broadly speaking, legislatures or courts could account for the platform nature of secondary copyright cases either by modifying the form of liability or by modifying remedies. Each will be considered in turn. Ultimately, however, reframing the issue has its own benefits apart from any specific doctrinal adjustments.

A. Implications for Liability

With respect to liability, as discussed above, platforms may need a relatively strong form of immunity to avoid incentives to over-block content, particularly when speech interests are at stake. The DMCA counter-notification provisions are aimed at addressing this issue,[70] but those provisions are relatively weak. There are technical requirements for an effective counter-notification, such that it demands some sophistication on the part of the user to take advantage of them.[71] There is a ten-day waiting period before the material can be put back.[72] Moreover, the counter-notification provision is itself an immunity provision and failure to comply is not on its own a basis for a claim against the intermediary.[73] Separately, the Ninth Circuit has held that copyright holders must “consider fair use before sending a takedown notification” as part of avoiding liability under § 512(f).[74] But § 512(f) is also weak insofar as it requires only “a subjective good faith belief that the use was not authorized by law.”[75]

The status quo has led to a certain amount of over-blocking of content—fair use content in particular.[76] Something stronger than § 512 may be needed to discourage this.[77] One possibility could be to provide an absolute immunity to intermediaries modeled on 47 U.S.C. § 230 for hosting anything other than an exact or near-exact copy of a copyrighted work. The goal would be to take cases like Lenz squarely outside of the realm of secondary liability. Such a scheme could be augmented with a requirement that copyright holders have a good faith belief that the subject of a notification is an exact or near-exact copy of their work to avoid liability under § 512(f).

To be sure, such a proposal is certainly over-inclusive. There will be many non-exact copies that are nevertheless infringing. That is in some sense, though, the point of the scheme. There would be no point in providing intermediary immunity for noninfringing works—in the absence of direct liability, there would be no secondary liability anyway. The lack of secondary liability for some infringing works is the tradeoff for trying not to incentivize the blocking of noninfringing works. What is needed, and what the line between exact and non-exact copies could provide, is a rule that is clear enough for a platform to feel comfortable knowing when its activities fall on the nonliability side of the line. Moreover, the experience under § 230 (and the private ordering discussed below) point to the fact that major sites will continue to block some content even when under no legal obligation to do so.[78]

Thus, there should potentially be some works as to which platforms are not secondarily liable even upon notice. Conversely, as to the rest, platforms should potentially bear some liability even if they have not received notice. Platforms could be required to take reasonable steps to detect and prevent infringement, with reasonableness measured by balancing the relative costs and benefits, as in the tort models mentioned above. Secondary trademark liability has moved in the direction of requiring such reasonable measures, despite the relevant standard being framed in terms of knowledge.[79] Secondary copyright liability could do the same, at least in that subset of copyright cases where free speech and chilling effects concerns are less pressing. On the flip side, the costliness or unreasonableness of certain measures should be seen as good reason not to require them, as with terminating Internet access, such as in the Cox cases.

As noted earlier, vicarious copyright liability does not fit the model of either paralleling secondary patent liability or hinging on knowledge or intent.[80] Thus, one might see in the existing standards for vicarious copyright liability a scheme for imposing liability beyond notice-based liability. This is true, but arguably existing vicarious copyright liability goes too far. The standard, which asks whether the defendant “has the right and ability to supervise the infringing activity and also has a direct financial interest in such activities,”[81] appears to contemplate liability even where the measures the defendant fails to take would be extremely costly. In that sense, vicarious copyright liability appears to be akin to a form of strict liability.

The strict liability nature of vicarious liability potentially makes sense in the agency-type cases from which the doctrine arose at common law.[82] An employer is liable for its employees’ torts not only because it may be able to supervise them efficiently, but also because it is appropriate to assign the responsibility to the employer as the party best able to bear the risk of such accidents.[83] Vicarious liability is thus an alternative form of responsibility-based liability. Contributory liability bases responsibility on knowledge or intent, while vicarious liability bases responsibility on the relationship between the parties. Neither is necessarily based on efficiency considerations. It may be that in the cases expanding vicarious copyright liability, imposing such liability on dance halls and flea markets is in fact efficient.[84] If so though, we ought to explicitly assess that efficiency and base liability on it, rather than on just the “right and ability to supervise” and “direct financial interest.”[85]

B. Implications for Remedies

An alternative way to account for the different roles of platforms and users is to adjust the remedies for infringement rather than to adjust liability. The traditional rule is that a plaintiff is free to collect the full measure of damages from either the direct infringer or a secondary infringer (but not both).[86] As explained above, however, treating platforms the same as their users can lead platforms to over-block content.[87] In the context of damages, this means that a damages regime optimized for direct liability is potentially excessive as applied to the platform instead. This may be particularly true for noncommercial platforms.

One way to address this would be to eliminate the possibility of statutory damages for platform liability. Some have already expressed concerns that statutory damages against direct infringers can be excessive.[88] Such concerns may be magnified in the context of platform liability. High statutory damages can unduly chill the development of the platforms themselves.[89] At a more micro level, high statutory damages also have the potential to cause over-blocking of specific pieces of content, at least insofar as the platform considers some content to be riskier than other content.

On the other hand, commercial platforms may sometimes actually make more money from infringing activity (through ad revenue, for example) than the direct infringers would have made on their own.[90] Thus, incentives can also be misaligned in the other direction, such that platforms may be incentivized to avoid looking for infringing content, absent a legal requirement to do so.

A misalignment of incentives in this direction might be best addressed through the disgorgement remedy.[91] The copyright statute provides, as a matter of course, for the disgorgement of the defendant’s profits.[92] Unlike, say, under trademark law,[93] disgorgement in copyright law does not depend on the willfulness or fault of the defendant. The ready availability of a disgorgement remedy is somewhat tempered by the limitation that the profits disgorged be “attributable to the infringement.”[94] That limitation, however, is counter-balanced by placing the burden of proof on the defendant to show what profits are not attributable to infringement.[95]

The net result of these rules is that in cases of non-exact direct copyright infringement, courts have struggled to determine how to apportion profits between the infringing and noninfringing portions of a work, often erring on the side of over-compensation.[96] When the plaintiff’s work is incorporated into some much larger work and the defendant’s work is then sold or monetized as a whole, it can be hard to know how to apportion the defendant’s profits.[97] Apportioning based on what percentage of the defendant’s work is infringing can either undershoot or overshoot the mark, depending on whether the infringing content is the heart of the work or merely an insignificant part of it.[98] Determining what the defendant would have made without using the infringing content can be an impossible hypothetical calculation.

Disgorgement in platform cases involving exact copies has the potential to be much more straightforward. Platforms like YouTube monetize individual pieces of content by showing advertisements.[99] That ad revenue can very reasonably be traced directly to the infringing content. Thus, even if in non-platform cases one might worry about excessive disgorgement awards, there may be less reason to worry in platform cases.

On the other hand, one might worry in the other direction that relying primarily on a disgorgement remedy against platforms could lead to under-deterrence. After all, it might seem that a defendant forced to disgorge profits is no worse off than if it had not infringed at all.[100] Defendants might therefore have an incentive to infringe as much as they can, knowing that they will be no worse off as to the infringements that are caught and better off as to the infringements that get by.

While it’s true, however, that a defendant forced to pay a disgorgement remedy is no worse off than if it had not infringed, the defendant in that situation is worse off than if it had gotten a license from the plaintiff. There is money to be made by the defendant’s activities. Rational parties will hopefully find some way to split that surplus, so that both parties are better off than if the activities had not occurred.[101] A defendant subject to a disgorgement remedy is getting nothing, while a licensed defendant is getting something. In this context, then, a disgorgement remedy alone can have a positive incentive effect.

C. Striking a Balance

In the end, the two main platform principles described above compete with each other. The principle of avoiding over-blocking counsels toward less platform liability. The principle of minimizing transaction costs will often counsel toward more. Depending on the weight one attaches to the interests served by each principle, one could advocate for different balances between them. These could range from applying § 230’s strong immunity to IP (or at least copyright) claims[102] to holding platforms strictly liable for the infringements they facilitate.[103] The specific proposals above, which fall in between those extremes, may or may not strike the right balance.

The key point is not so much whether one agrees with any of these specific proposals, but rather to reframe the analysis and design choices around these platform principles instead of a focus on knowledge and intent. Doing so could be more likely to serve the relevant interests. There has been some recent interest in amending § 512, for example.[104] The framework above helps to explain why certain aspects of § 512, such as the standard for “red flag” knowledge and the relationship between § 512 and vicarious liability, have been so messy and contentious.[105] Trying to apply non-platform principles to platform cases is trying to fit square pegs into round holes. Rethinking § 512 using platform principles could provide a better basis for legislative change.

One might ask what’s at stake in reformulating the substantive law of secondary copyright liability, given that so much of what happens on platforms is now driven by private ordering rather than the precise contours of the law.[106] YouTube’s Content ID system, for example, does more than what the law requires by engaging in ex ante filtering, and it provides for possibilities, such as monetization, that are not directly contemplated by the applicable law.[107] Yet the bargaining between rights holders and platforms that results in systems like Content ID occurs in the shadow of existing copyright law, and in particular in the shadow of the DMCA safe harbors.[108] Adjusting those safe harbors, for example, to provide complete immunity for non-exact copies, could influence the design of automated enforcement systems and perhaps, for example, diminish the extent to which fair use content is automatically blocked.

As described above, it also seems that courts themselves are managing to adapt legal standards to platform cases by reinterpreting terms like “knowledge.”[109] Again, courts may reach many of the right outcomes, but doing so in a more principled way will help to both justify those outcomes and determine how far to take them. All told, between judicial interpretation and private ordering, the status quo could well be reasonable or the best we can manage. Even then, platform principles would help to better explain why that is the case.

V. Conclusion

Platform cases are different from non-platform cases. Using secondary liability law developed in the non-platform context to resolve platform cases is very likely to be suboptimal. In many ways, courts have acted on this intuition without even realizing it. Being more explicit about what makes for a good platform law helps to explain and justify certain aspects of existing secondary copyright law, while also suggesting how courts and Congress might improve the law in the future.


  1. 35 U.S.C. § 271(a) (“Except as otherwise provided in this title, whoever without authority makes, uses, offers to sell, or sells any patented invention, within the United States or imports into the United States any patented invention during the term of the patent therefor, infringes the patent.”).

  2. Id. § 271(b)–(c).

  3. See also Laura A. Heymann, Knowing How to Know: Secondary Liability for Speech in Copyright Law, 55 Wake Forest L. Rev. 333, 347–49 (2020) (explaining the ambiguities in what it means to have knowledge of infringement).

  4. See § 271(c).

  5. Aro Mfg. v. Convertible Top Replacement Co., 377 U.S. 476, 488 (1964).

  6. Global-Tech Appliances, Inc. v. SEB S.A., 563 U.S. 754, 765–66 (2011). In Commil USA, LLC v. Cisco Systems, the Court added the additional gloss that, while a good-faith belief that the induced conduct is noninfringing would negate the required mens rea under § 271(b), a good-faith belief that the plaintiff’s patent is invalid does not. Commil USA, LLC v. Cisco Sys., 575 U.S. 632, 641–42 (2015).

  7. A&M Recs., Inc. v. Napster, Inc., 239 F.3d 1004, 1019 (9th Cir. 2001) (quoting Gershwin Publ’g Corp. v. Columbia Artists Mgmt., 443 F.2d 1159, 1162 (2d Cir. 1971)).

  8. Id. at 1021.

  9. Sony Corp. of Am. v. Universal City Studios, Inc., 464 U.S. 417, 442 (1984).

  10. Metro-Goldwyn-Mayer Studios Inc. v. Grokster, Ltd., 545 U.S. 913, 930, 936–37 (2005).

  11. Id. at 934–35.

  12. 17 U.S.C. § 512(c)(1)(A).

  13. Sony, 464 U.S. at 439.

  14. Grokster, 545 U.S. at 936.

  15. See, e.g., id. at 930 (“[D]octrines of secondary liability emerged from common law principles and are well established in the law . . . .”). In this regard, it is, perhaps, surprising that the Court has disavowed a link between secondary trademark liability and secondary copyright liability. Sony, 464 U.S. at 439 n.19 (“We have consistently rejected the proposition that a similar kinship exists between copyright law and trademark law, and in the process of doing so have recognized the basic similarities between copyrights and patents. Given the fundamental differences between copyright law and trademark law, in this copyright case we do not look to the standard for contributory infringement set forth in Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U.S. 844, 854–55 (1982), which was crafted for application in trademark cases.” (citations omitted)).

  16. See Sentius Int’l, LLC v. Apple Inc., No. 4:20-cv-00477-YGR, 2020 WL 6081775, at *5, *7 & n.3 (N.D. Cal. Oct. 15, 2020) (“Contributory infringement is expressly defined in 35 U.S.C. § 271(c) and requires knowledge of the patent; vicarious infringement stems from common law principles and does not.”).

  17. Akamai Techs., Inc. v. Limelight Networks, Inc., 797 F.3d 1020, 1022 (Fed. Cir. 2015) (en banc) (per curiam). The Federal Circuit in Akamai recognized that the “use of the term ‘vicarious liability’” in the context of divided infringement “is a misnomer.” Id. at 1022 n.2. Divided infringement is an issue of direct, rather than secondary, liability. See id. at 1022.

  18. A&M Recs., Inc. v. Napster, Inc., 239 F.3d 1004, 1022 (9th Cir. 2001) (quoting Gershwin Publ’g Corp. v. Columbia Artists Mgmt., 443 F.2d 1159, 1162 (2d Cir. 1971)).

  19. See 17 U.S.C. § 512(c)(1)(B) (requiring that the service provider “does not receive a financial benefit directly attributable to the infringing activity, in a case in which the service provider has the right and ability to control such activity”); see also Viacom Int’l, Inc. v. YouTube, Inc., 676 F.3d 19, 36–38 (2d Cir. 2012) (rejecting the argument that “the control provision [in § 512(c)(1)(B)] codifies the common law doctrine of vicarious copyright liability”).

  20. See Aro Mfg. v. Convertible Top Replacement Co., 377 U.S. 476, 478–79 (1964).

  21. See id. at 500–02 (“[A] contributory infringer is a species of joint-tortfeasor, who is held liable because he has contributed with another to the causing of a single harm to the plaintiff.”).

  22. See, e.g., UMG Recordings, Inc. v. Shelter Cap. Partners LLC, 718 F.3d 1006, 1013 (9th Cir. 2013).

  23. See generally Matthew Sag, Copyright Safety for Generative AI, 61 Hous. L. Rev. 295 (2023). If a particular prompt causes a generative AI system to output an infringing work, it is not clear whether the user or the system (or perhaps both) should be considered the primary infringer. The usual “volition” test for direct copyright infringement seems somewhat indeterminate. See Religious Tech. Ctr. v. Netcom On-Line Comm’n Servs., 907 F. Supp. 1361, 1370 (N.D. Cal. 1995) (requiring “some element of volition or causation” for direct copyright liability). In any event, whichever party is not a primary infringer could then potentially be held liable as a secondary infringer. In this context, neither party could be said to be acting as a platform for the other. In particular, the generative AI system seems far too involved in the actual creation of the infringing content to be regarded as a mere platform for the user’s content.

  24. See Peter S. Menell & David Nimmer, Unwinding Sony, 95 Calif. L. Rev. 941, 985 (2007) (arguing that there was little reason to borrow from patent law in deciding the Sony case).

  25. See 35 U.S.C. § 271(c).

  26. See Sony Corp. of Am. v. Universal City Studios, Inc., 464 U.S. 417, 442–43 (1984).

  27. See A&M Recs., Inc. v. Napster, Inc., 239 F.3d 1004, 1020 (9th Cir. 2001) (“We observe that Napster’s actual, specific knowledge of direct infringement renders Sony’s holding of limited assistance to Napster.”).

  28. See Global-Tech Appliances, Inc. v. SEB S.A., 563 U.S. 754, 758, 768 (2011).

  29. Metro-Goldwyn-Mayer Studios Inc. v. Grokster, Ltd., 545 U.S. 913, 936–37 (2005) (emphasis added).

  30. Id. at 940 & n.13.

  31. Id. at 940.

  32. Id.

  33. Compare 35 U.S.C. § 271(c) (imposing liability on one who acts “knowing the same to be especially made or especially adapted for use in an infringement of such patent”), with Gershwin Publ’g Corp. v. Columbia Artists Mgmt. Inc., 443 F.2d 1159, 1162 (2d Cir. 1971) (imposing liability on one who acts “with knowledge of the infringing activity”).

  34. See Aro Mfg. v. Convertible Top Replacement Co., 377 U.S. 476, 488 (1964).

  35. See A&M Recs., Inc. v. Napster, Inc., 239 F.3d 1004, 1020 (9th Cir. 2001) (“Contributory liability requires that the secondary infringer ‘know or have reason to know’ of direct infringement.” (citations omitted)).

  36. See Tiffany (NJ) Inc. v. eBay Inc., 600 F.3d 93, 103–06 (2d Cir. 2010).

  37. See Inwood Lab’ys, Inc. v. Ives Lab’ys, Inc., 456 U.S. 844, 846 (1982).

  38. Id. at 854.

  39. See id. (asking whether a defendant “intentionally induces another to infringe” or whether a defendant “knows or has reason to know” that the other entity is infringing).

  40. An initial question is whether the Inwood standard applies to platforms at all, given the reference to the defendant “supply[ing a] product.” Id. Courts have generally interpreted the term “product” in the Inwood standard broadly and applied the standard to platforms and markets, both online and off. See Hard Rock Cafe Licensing Corp. v. Concession Servs., 955 F.2d 1143, 1148–49 (7th Cir. 1992).

  41. Tiffany, 600 F.3d at 107.

  42. See id. at 108 (quoting the Supreme Court in the Sony case describing the Inwood standard as a “narrow standard”).

  43. See David H. Bernstein & Michael R. Potenza, Why the Reasonable Anticipation Standard Is the Reasonable Way to Assess Contributory Trademark Liability in the Online Marketplace, 2011 Stan. Tech. L. Rev., no. 9, at 15–16.

  44. Luxottica Grp., S.P.A. v. Airport Mini Mall, LLC, 932 F.3d 1303, 1314–15 (11th Cir. 2019).

  45. Id. at 1314.

  46. See Felix T. Wu, Collateral Censorship and the Limits of Intermediary Immunity, 87 Notre Dame L. Rev. 293, 308 (2011).

  47. Id.

  48. Id. at 306–07.

  49. See Alfred C. Yen, Torts and the Construction of Inducement and Contributory Liability in Amazon and Visa, 32 Colum. J.L. & Arts 513, 529 (2009) (arguing that courts model secondary copyright liability too much after intentional torts rather than negligence and thereby fail to account for the reasonableness and “the potential unintended consequences of the proposed precaution”); see also Mark P. McKenna, Probabilistic Knowledge of Third-Party Trademark Infringement, 2011 Stan. Tech. L. Rev., no. 10, at 10 (making a similar argument with respect to secondary trademark liability); Heymann, supra note 3, at 377 (suggesting a move toward “the negligence aspects of the [secondary liability] framework” as a way to deal with the difficulties in defining and assessing knowledge).

  50. See 47 U.S.C. § 230(c)(1) (“No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.”). Section 230 does not apply to intellectual property claims. See § 230(e)(2) (“Nothing in this section shall be construed to limit or expand any law pertaining to intellectual property.”).

  51. See Zeran v. Am. Online, Inc., 129 F.3d 327, 333 (4th Cir. 1997); see also, e.g., Universal Commc’n Sys. v. Lycos, Inc., 478 F.3d 413, 420 (1st Cir. 2007) (“We . . . join the other courts that have held that Section 230 immunity applies even after notice of the potentially unlawful nature of the third-party content.”).

  52. See Zeran, 129 F.3d at 333 (“Because service providers would be subject to liability only for the publication of information, and not for its removal, they would have a natural incentive simply to remove messages upon notification, whether the contents were defamatory or not. Thus, like strict liability, liability upon notice has a chilling effect on the freedom of Internet speech.” (citation omitted)).

  53. See Douglas Lichtman & William Landes, Indirect Liability for Copyright Infringement: An Economic Perspective, 16 Harv. J.L. & Tech. 395, 397 (2003); Tun-Jen Chiang, The Conduit Theory of Secondary Liability in Patent and Copyright Law, 23 Nev. L.J. 65, 84–86 (2022).

  54. Lichtman & Landes, supra note 53.

  55. Id. at 399.

  56. See BMG Rts. Mgmt. (US) LLC v. Cox Commc’ns, Inc., 881 F.3d 293, 310–11 (4th Cir. 2018); Sony Music Ent. v. Cox Commc’ns, Inc., 464 F. Supp. 3d 795, 815–16 (E.D. Va. 2020).

  57. See BMG, 881 F.3d at 298–99.

  58. See Sony, 464 F. Supp. 3d at 816; BMG Rts. Mgmt. (US) LLC v. Cox Commc’ns, Inc., 199 F. Supp. 3d 958, 980 (E.D. Va. 2016). The district court also found, and the Fourth Circuit upheld, that Cox was not entitled to the safe harbor under § 512(a) because it had failed to reasonably implement a policy of terminating repeat infringers as required under § 512(i). See BMG, 881 F.3d at 303–05.

  59. BMG, 199 F. Supp. 3d at 979; Sony, 464 F. Supp. 3d at 816.

  60. See BMG, 199 F. Supp. 3d at 978–79; Sony Music Ent. v. Cox Commc’ns, Inc., 426 F. Supp. 3d 217, 232–33 (E.D. Va. 2019).

  61. See BMG, 881 F.3d at 310–11.

  62. See, e.g., BMG, 199 F. Supp. 3d at 970–71.

  63. See Yen, supra note 49, at 527 (noting presciently that “[i]f courts imposed liability on the power company (or any other provider of basic services) simply for continuing to provide services to customers accused of infringement, socially undesirable results would follow”).

  64. See Aro Mfg. v. Convertible Top Replacement Co., 377 U.S. 476, 500 (1964) (“[A] contributory infringer is a species of joint-tortfeasor, who is held liable because he has contributed with another to the causing of a single harm to the plaintiff.”).

  65. See, e.g., Posecai v. Wal-Mart Stores, Inc., 752 So.2d 762, 766 (La. 1999) (imposing on business owners “a duty to implement reasonable measures to protect their patrons from criminal acts when those acts are foreseeable”).

  66. See, e.g., Calmes v. Goodyear Tire & Rubber Co., 575 N.E.2d 416, 421 (Ohio 1991) (explaining that a products liability claim is viable even when the product has been “unreasonably misuse[d]” so long as the misuse is not unforeseeable (alteration in original)).

  67. See, e.g., Vince v. Wilson, 561 A.2d 103, 105 (Vt. 1989) (holding that the tort of negligent entrustment applies to anyone who supplies the chattel in question even if they do not own or control it).

  68. See Posecai, 752 So.2d at 766.

  69. See McKenna, supra note 49, at 8 (making a similar argument with respect to secondary trademark liability).

  70. See 17 U.S.C. § 512(g).

  71. See id. § 512(g)(3).

  72. See id. § 512(g)(2)(C) (requiring material to be put back “not less than 10, nor more than 14, business days following receipt of the counter notice”).

  73. See id. § 512(g)(1) (“Subject to paragraph (2), a service provider shall not be liable to any person for any claim based on the service provider’s good faith disabling of access to, or removal of, material or activity claimed to be infringing . . . .”).

  74. Lenz v. Universal Music Corp., 815 F.3d 1145, 1148 (9th Cir. 2016); see also § 512(f)(1) (imposing liability on “[a]ny person who knowingly materially misrepresents under this section . . . that material or activity is infringing”).

  75. Lenz, 815 F.3d at 1148.

  76. See Jennifer M. Urban et al., Notice and Takedown in Everyday Practice 95–96 (ver. 2, Mar. 2017), https://osf.io/preprints/socarxiv/59m86 [https://perma.cc/CTD9-TMA4]; see also Xiyin Tang, Privatizing Copyright, 121 Mich. L. Rev. 753, 774–75 (2023) (discussing the failure of private agreements between platforms and copyright holders to account for fair use).

  77. Tun-Jen Chiang argues that existing secondary liability law (both patent and copyright) reflects “an absolute prohibition on collateral harms.” Chiang, supra note 53 at 89. By this he seems to mean primarily that the law avoids imposing liability when doing so would directly impede lawful uses, as with restricting dual-use technologies. My focus here is on a broader concept of collateral consequences that includes over-blocking not as a result of a legal mandate but out of fear of legal liability. That form of “collateral harm” is not precluded by existing secondary copyright law. Chiang suggests that the “substantial non-infringing uses” exception eliminates or minimizes any chilling effects from contributory liability. See id. at 96, 100. While this might generally be true for patent law, it is not true for copyright law, see supra note 76 and accompanying text, showing again how secondary patent cases and secondary copyright cases are structurally different.

  78. See, e.g., Kate Klonick, The New Governors: The People, Rules, and Processes Governing Online Speech, 131 Harv. L. Rev. 1598, 1625–30 (2018).

  79. See supra text accompanying notes 44–45; see also Coach, Inc. v. Goodfellow, 717 F.3d 498, 504 (6th Cir. 2013) (finding the defendant to be contributorily liable where he failed to “take . . . reasonable measures to prevent use of flea market resources for unlawful purposes, and failed even to undertake a reasonable investigation”).

  80. See supra notes 16–19 and accompanying text.

  81. Gershwin Publ’g Corp. v. Columbia Artists Mgmt., 443 F.2d 1159, 1162 (2d Cir. 1971).

  82. See, e.g., Hinman v. Westinghouse Elec. Co., 471 P.2d 988, 990 (Cal. 1970) (explaining that employers are vicariously liable for their employees’ torts even when the employer is not at fault).

  83. See id. (“[T]he modern and proper basis of vicarious liability of the [employer] is not his control or fault but the risks incident to his enterprise.”).

  84. See, e.g., Fonovisa, Inc. v. Cherry Auction, Inc., 76 F.3d 259, 263–64 (9th Cir. 1996) (finding that the plaintiffs stated a claim for vicarious copyright infringement against a flea market, at least in part by analogy to claims against dance halls).

  85. Alternatively, flea markets may have a sufficiently close relationship to their vendors and dance halls to the bands that play there in order to justify a relationship-based form of vicarious liability. That too is not captured by the elements of “right and ability to supervise” and “direct financial interest,” at least as those terms have been interpreted. See A&M Recs., Inc. v. Napster, 239 F.3d 1004, 1022–24 (9th Cir. 2001) (finding vicarious liability despite Napster and its users having nothing more than an arms-length provider-user relationship).

  86. See Aro Mfg. v. Convertible Top Replacement Co., 377 U.S. 476, 502–03 (1964).

  87. See supra notes 46–48 and accompanying text.

  88. See Pamela Samuelson & Tara Wheatland, Statutory Damages in Copyright Law: A Remedy in Need of Reform, 51 Wm. & Mary L. Rev. 439, 441 (2009) (“Awards of statutory damages are frequently arbitrary, inconsistent, unprincipled, and sometimes grossly excessive.”); Oren Bracha & Talha Syed, The Wrongs of Copyright’s Statutory Damages, 98 Tex. L. Rev. 1219, 1220 (2020) (“On its face, there seems to be something fundamentally wrong with an individual of ordinary wealth having to pay over a million dollars for sharing twenty-four copyrighted sound recordings online.”).

  89. See Stephanie Berg, Remedying the Statutory Damages Remedy for Secondary Copyright Infringement Liability: Balancing Copyright and Innovation in the Digital Age, 56 J. Copyright Soc’y USA 265, 312 (2008); Michael A. Carrier, Innovation for the 21st Century: Harnessing the Power of Intellectual Property and Antitrust Law 160–61 (2009).

  90. See, e.g., Monetization Policies, YouTube, https://www.youtube.com/howyoutubeworks/policies/monetization-policies/ [https://perma.cc/T3WM-35SL] (last visited Aug. 30, 2023); Channel Monetization Policies, YouTube, https://support.google.com/youtube/answer/1311392 [https://perma.cc/BPD5-PZLB] (last visited Aug. 30, 2023).

  91. See Pamela Samuelson et al., Recalibrating the Disgorgement Remedy in Intellectual Property Cases, 100 B.U. L. Rev. 1999, 2029 (2020) (discussing the role of disgorgement in achieving an appropriate level of deterrence).

  92. See 17 U.S.C. § 504(b) (“The copyright owner is entitled to recover the actual damages suffered by him or her as a result of the infringement, and any profits of the infringer that are attributable to the infringement and are not taken into account in computing the actual damages.”).

  93. See 15 U.S.C. § 1117(a) (permitting recovery of the defendant’s profits for “a willful violation under section 1125(c) of this title”); Romag Fasteners, Inc. v. Fossil, Inc., 140 S. Ct. 1492, 1497 (2020) (rejecting a requirement to show willfulness in order to obtain the defendant’s profits in an ordinary trademark infringement case, but nevertheless affirming that “a trademark defendant’s mental state is a highly important consideration in determining whether an award of profits is appropriate”).

  94. 17 U.S.C. § 504(b).

  95. See id. (“In establishing the infringer’s profits, the copyright owner is required to present proof only of the infringer’s gross revenue, and the infringer is required to prove his or her deductible expenses and the elements of profit attributable to factors other than the copyrighted work.”).

  96. See Samuelson et al., supra note 91, at 2050–52.

  97. See, e.g., Frank Music Corp. v. Metro-Goldwyn-Mayer, Inc., 772 F.2d 505, 510 (9th Cir. 1985) (defendant’s Las Vegas show had ten acts, only one of which infringed plaintiff’s copyright).

  98. See Samuelson et al., supra note 91, at 2052–54.

  99. See sources cited supra note 90.

  100. See Bert I. Huang, The Equipoise Effect, 116 Colum. L. Rev. 1595, 1598 (2016).

  101. See Chiang, supra note 53, at 84 (noting that the ultimate goal of intellectual property law is for rightsholders to be appropriately paid for uses of their IP, not to stop those uses from occurring).

  102. Cf. Mark A. Lemley, Rationalizing Internet Safe Harbors, 6 J. on Telecomm. & High Tech. L. 101, 102, 112–15 (2007) (arguing for a “uniform safe harbor rule” across both IP and non-IP claims, although one that is ultimately less protective than § 230 while being more protective than the DMCA).

  103. As noted above, vicarious copyright liability potentially functions as a form of strict liability. See supra note 81 and accompanying text.

  104. See U.S. Copyright Office, Section 512 of Title 17: A Report of the Register of Copyrights 1–2 (May 2020), https://www.copyright.gov/policy/section512/section-512-full-report.pdf [https://perma.cc/T3JU-VQY9].

  105. See id. at 110–36.

  106. See Matthew Sag, Internet Safe Harbors and the Transformation of Copyright Law, 93 Notre Dame L. Rev. 499, 538 (2017).

  107. See id. at 541–42.

  108. See id. at 541 (suggesting that “YouTube’s development of Content ID appears to have been spurred by the Viacom litigation,” in which Viacom argued for narrow interpretations of the DMCA safe harbors).

  109. See supra notes 33–35 and accompanying text.