Once upon a time, “intellectual property” wasn’t even a concept. The term has roots in early federal court cases but came into vogue only in the later twentieth century—perhaps because it was a lot cooler-sounding to tell people you “did intellectual property” than that you were a “copyright lawyer.” However, with the emergence of the Internet at the end of the 1900s, the cost of creating and distributing works plummeted, resulting in both the disruption of content industries and the generation of countless new avenues for expression. Copyright is no longer a body of law that solely governs the production of copies of works of authorship. It now touches every aspect of our daily lives—not only the entertainment and other content we consume but also the architecture that makes up the digital means by which we consume them.
The 19th IPIL National Conference, held in Santa Fe, New Mexico on June 3, 2023, featured an array of presentations that reflected the broad reach of present-day copyright law. The assembled scholars and commentators investigated how a law last revised in 1976 could be read to govern internet platforms and provide some constraint on the dizzying growth of artificial intelligence. They looked inward to ask how insights from social science could improve the substantial similarity doctrine and looked outward to ask how fair use could better serve social justice. And they also reminded us that copyright still governs traditional creative expression via the scènes à faire doctrine.
To begin, Christopher Buccafusco and Rebecca Tushnet explore what behavioral science can tell us about the copyright doctrine in “Of Bass Notes and Base Rates: Avoiding Mistaken Inferences About Copying.” The authors focus on the process of proving infringement, which requires (inter alia) that plaintiffs show that the defendant copied their work. Where works—especially musical works—are similar but not identical, plaintiffs often seek to prove copying occurred by retaining experts to testify that the degree of similarity between their work and the accused work is so high that the only valid inference is that the defendant copied. The authors show that this practice suffers from a fatal flaw: base-rate error. Their argument assumes that the experts know how likely it is that two given songs will exhibit similarity absent any copying—i.e., what the base rate of similarity is. This simple and elegant insight has major implications for copyright law and practice. In particular, the authors show that the widespread practice of retaining experts to infer copying based on the degree of similarity between two works is unsound and that it should be deemed inadmissible under the Federal Rules of Evidence.
Betsy Rosenblatt explores the implications of a different idea exogenous to copyright in her article, “Considering the Role of Fairness in Copyright Fair Use.” Rosenblatt invites us to consider whether the fair use doctrine truly promotes fairness and whether that doctrine should consider the relative status of owners and infringers. Her thought experiment isolates several upsides that may follow if copyright treated unauthorized uses by less privileged individuals more leniently, including allowing such individuals to push back against dominant culture and its norms, allowing them more pathways to self-expression and success, and deterring more privileged speakers from engaging in cultural appropriation. In doing so, Rosenblatt not only suggests an ambitious and more inclusive vision of fair use but adds to the growing literature on how copyright can promote social justice.
On a different note, a major convergence of copyright and technology is the emergence of artificial intelligence and, in particular, large language models (LLMs) that generate high-quality text and images based solely on a prompt. In “Copyright Safety for Generative AI,” Matthew Sag examines how the fair use doctrine applies in this novel context. LLMs’ ability to generate content results from a process of “learning”: the decomposition and abstraction of trillions of data points that include copyrighted works of authorship. Professor Sag first argues that LLMs’ use of these data points is noninfringing fair use because LLMs do not memorize content from the works themselves, but rather incorporate latent qualities of the training data. However, he then cautions that some text-to-image models may memorize works under certain conditions, which are particularly salient in the context of copyrighted characters. Sag concludes by suggesting a set of guidelines that promise to mediate between the freedom to use copyrighted works as training data while avoiding the possibility that the outputs of LLMs will infringe owners’ rights to those works.
Zahr Said’s “Grounding the scènes à faire Doctrine” provides a rich and incisive critique of one of the most familiar and elusive doctrines in all of copyright. The idea of the scènes à faire doctrine is that there are certain tropes that are so common that they cannot be the subject of copyright protection. Professor Said identifies scènes à faire as one of several limiting doctrines intended to prevent the copyright monopoly from suppressing future expression. Her analysis also reveals that this doctrine is often conflated with other limiting doctrines, such as merger and idea/expression, thereby limiting its efficacy. Said proceeds to disentangle scènes à faire from similar, adjacent doctrines to reveal its pragmatic potential. So isolated, Said shows that scènes à faire has the capacity to contribute specific principles to copyright infringement analysis, imbuing with more rigor an analysis long thought to be hopelessly indeterminate.
And finally, a look at copyright and social platforms. Unauthorized uses of works of authorship proliferate on social media websites that facilitate their reproduction and display. But, as Felix Wu points out in “The Structure of Secondary Copyright Liability,” the Copyright Act does not have provisions establishing secondary liability for encouraging or facilitating infringement. However, the Patent Act, Professor Wu points out, does have such provisions, and courts have fashioned copyright’s law of secondary liability modeling after those provisions of patent law. The problem with this is that patent law does not typically occur at the platform level, as copyright infringement does. As a result, Wu argues, copyright’s law of secondary liability focuses too much on mens rea and fault and not enough on under- and over-deterrence. To remedy this problem, Wu suggests adjusting copyright’s law of secondary liability in terms of both substance (limiting liability to verbatim reproduction) and remedies (eliminating statutory damages for secondary infringement). These changes promise to better balance the promise of recruiting platforms to police infringement and avoid chilling of noninfringing online conduct.
Copyright has come a long way from the 1790 Act, when it was largely a doctrine regulating the printing of texts via administrative formalities. The domain of copyright has become as varied and capacious as the creativity and technology it governs, illustrated by the range and divergence of the contributions of each of this year’s authors at the IPIL National Conference.