George Mason University Antonin Scalia Law School

Authors are Humans and Creativity is a Function of Humanness: What the Mannion Court Can Teach Us About Generative AI’s Relationship to Authorship

By Molly Stech*

* The blog post below and the law review article it links to are the individual thoughts and views of the author and should not be attributed to any entity with which she is currently or has been affiliated.

silver copyright symbolDespite a recent decision from the Beijing Internet Court, there is growing consensus that artificial intelligence (AI) can be used as a tool, but that a human author must have ideated a copyrighted work and that the resultant creative work is the outcome of that person’s intellect and personality. Despite some international convergence on this issue, however, it is worth reviewing the backdrop of this issue and uncovering some of the more vexing practicalities regarding the level of creative autonomy a person must exercise to receive a copyright registration. The threshold for creativity in copyright law is low across jurisdictions, but how low is it? Are fifty binary choices enough to confer authorship? Are 624 generative AI prompts enough? Similar to other areas of copyright, such as the idea-expression dichotomy, or the unpredictability of the U.S. fair use doctrine, there are almost no bright lines to be drawn in the context of AI.

In a forthcoming paper, I review the law and the jurisprudential landscape on AI “authorship,” as well as academic commentary on the topic, and conclude that the bedrock principles of copyright law would not be served by permitting an acknowledgment of an AI system or algorithm as an author. Although AI is a groundbreaking, even revolutionary, technology, the ways in which it challenges the traditional contours of copyright law are not entirely new. We know from the New York Bridgeman decision in 1999 that skill and labor – and even creativity – in the production stage of creativity are meaningless unless the output (or in copyright parlance, the “work”) exhibit creativity. The Court of Justice of the European Union (CJEU) and United States courts point to an “author’s own intellectual creation” and a “modicum of creativity,” respectively, in ascertaining whether something merits copyright protection. As the European Union implements the world’s first AI Act, and as the U.S. Copyright Office reviews applications for AI-assisted works, underscoring the importance of human authorship is paramount to ensuring laws and courts are well-equipped with the rationale underlying the important distinction between human creativity and machine-generated outputs.

Human authorship has always been, and continues to be, a foundational requirement for copyright protection to subsist in a work. AI challenges this prerequisite but does not overcome it. The output of generative AI is not discernibly different from the output of a human author and therefore benefits from a false sheen of originality. While some argue that prompt engineering fulfills the requirements of originality––as noted above, the threshold for originality is quite low across jurisdictions––prompting still lacks the requisite link between human creativity and the resulting work to receive copyright protection. International copyright treaties and domestic copyright law must be interpreted as aiming to provide copyright’s exclusive rights to works that reflect human originality and that reward human beings. A 2006 New York district court case outlined three means by which photographs can demonstrate originality: rendition, timing, and creation of the subject. My paper proposes that each of these mechanisms, understood through the prism of generative AI, remains applicable for analyzing whether human originality subsists in a given work. Originality exists along a sliding scale, resulting in a mix of thin copyrights and medium copyrights and thick copyrights. While it may not always be the case as the technology evolves, the current relationship between generative AI and its user results in outputs that are generally too detached from the user’s creativity to satisfy the requirements of copyrightable authorship. Generative AI remixes the content on which it has been trained according to its algorithm and prompts. Copyright protection is a privilege and it can only be earned by humans by way of their own intellectual creations.