AI Authorship: A Case of History Repeating Itself?

Abstract

The idea of computers creating works which might be copyrightable (be it called AI, computational creativity, algorithmic authorship, or anything else) is older than we tend to think, as are attempts to register such works for copyright. This project looks back at the history of registering computer generated works along with a number of other ideas about authorship which push the meanings of “authorship” in different directions and proposes some lessons we can learn.  In the mid-1950s, experiments in authoring music by computer began, followed by graphical works and compilations by the mid-1960s. Repeated attempts to register these works led to a 1965 policy statement from the Copyright Office that “[t]he crucial question is whether the work is one of human authorship with the computer merely being the instrument, or whether the creation is conceived and executed not by man but by a machine.” This position remains the position of the office, and is quoted in the Office’s 2023 Policy Statement on works containing AI-generated material.

I have tracked down what I believe to be the works in question and other works where copyright registration was attempted for a AI work. In addition, although some precedents like photographs and animal authorship are well-known, I also trace the nearly parallel issue of the registration of aleatory music – music where the expression is intentionally indeterminate. Barely explored in the literature, I believe the historical record which exists on this issue offers us a solution and insight.  In addition to practical solution, I believe that exploring the history of copyright registration of AI works also offers us insight into what exactly the purpose and nature of copyright is, with the focus on human authorship not merely being a legal requirement but central to copyright itself.