In this essay, Abraham Drassinower updates the argument in his 2015 book What’s Wrong With Copying (Harvard University Press) with an elaboration of the nature of copyright’s public domain: it is a domain of unauthorized yet lawful copying. As in his book, Drassinower explains that the public domain is not properly understood as a problem of balancing between copyright’s public benefits and private rewards. Instead, he understands the public domain as part of authorship and thus inseparable from copyright’s construction.
The elaboration in this essay is his introduction of readership to the understanding of copyright’s public domain. As he says in its opening pages, “I will argue that the work of authorship in copyright law and doctrine is radically inseparable from what we may call the work of readership. Setting forth the immanence of readership in authorship will help construe the public domain as a necessary presupposition of, rather than a constraint on, authorship.” (P. 3.)
Before outlining Drassinower’s argument, I want to explain why I love this essay:
First, Drassinower writes like a logician, building an intricate argument from very basic premises and brings the reader along step by step.
Second, despite its sophistication, his writing is conversational and intimate, as if he’s sitting and speaking with his reader.
Third, he does not import any methodology external to law to make his argument – e.g., economics, history, literature, sociology – deriving his argument from within copyright doctrine itself.
Fourth, although his book was published in 2015, this essay reminded me of the book’s timelessness, which is remarkable in a copyright field evolving rapidly with technological evolution.
Fifth and finally, this essay helped me think about the on-going AI and copyright debates in a new way – E.g, Are the inputs to gen-AI models infringing uses? — which I appreciate because I think we need some new ways of thinking about these problems.
Drassinower’s main contribution in this essay is to think about copyright subject matter as always already involving the public – the author’s audience – and thus centering readership as essential to copyright’s purpose. He says that although copyright is about authors, it is centrally about “the relation between authors and publics, creators and users” (P. 5) because an original work of expression (copyright’s subject matter) is addressed to an audience. He quotes Jorge Luis Borges to make the point clear that authors are not creating objects or things when producing an original work of expression, but initiating a process between people. “Picking up a book and opening it holds the possibility of an aesthetic happening. What are the words resting in a book? What are those dead symbols? Absolutely nothing. What is a book if we do not open it? It is simply a cube of paper and leather … but if we read it something strange happens, I believe it changes every time.” (P. 25.) This view, for Drassinower, follows from copyright’s foundational idea/expression dichotomy, which allows all authors to repeat and build on the ideas of others, facilitating and proliferating authorship through response.
Drassinower starts with what he admits is the obvious copyright proposition that “lawful copying is part and parcel of copyright” (P. 9) because not all copying is infringement. From there, however, he explains that because the kind of copying that is lawful includes criticism, parody, research, and the free availability of ideas, the immanence of readers and responses to authorship follows directly. Reading and authorial response is built into copyright law, as a bilateral construct to authorship and original works of expression. Readership in this way is part of copyright’s public “derived as an unfolding of the fundamentals of authorship itself. As a mode of public address, authorship necessarily implicates the audience it addresses. It contemplates responses.” (P. 12.)
What does this have to do with the copyright debates concerning generative AI? First, the AI debates are often (and most recently) about the copyrighted work as a thing which when copied to feed a large language model (LLM) is contested as an unlawful use. Second, the AI debates are often (and again most recently) about the harm that LLMs might cause authors through their capacity for market substitution. These are usually questions of the idea/expression dichotomy and fair use, and to my mind not hard questions at that. But new court decisions (linked above) have muddied these doctrinal waters, presumably because generative AI feels like an existential threat to many authors (and publishers). Drassinower’s essay clarifies these issues in novel ways.
His essay on the “work of readership” reminds us that these copyright and AI debates misunderstand copyright subject matter as an object rather than a relation, a product rather than an intersubjective process. A work of authorship creates an address – an engagement – as a shared activity. And it is that activity that must be “protected” by law – the thinking activity that demands a response – not the thing copied (to be or not to be ingested by a machine).
By this logic, technical uses (as tools) are not the kind of use copyright cares about because they have nothing to do with the author as speaker (or reader as responder). And personal uses or private copies likewise have nothing to do with copyright because “copying for personal use has nothing to do with addressing another.” (P. 33.) This means copies made and not distributed are also part of copyright’s public domain. Taken seriously, these two propositions would render lawful the copying done to make LLM models produced even through unauthorized downloads of millions of books and even if (perhaps especially if!) those LLM models can produce works that readers prefer to read.
I found Drassinower’s essay provocative in the way controversial and novel frameworks often are. Much like his book on which this essay builds, I predict there is more to mine in this essay, and I hope people read it as it was meant to provoke a response!
