From Player Piano to Generative AI: Artificial Expression and the Law of Copyright

American copyright law has experienced challenges and sometimes faltered when confronting mechanical reproduction. This Article retraces the rise and diffusion of the player piano and the phonograph, examining how they reshaped the making and listening to music. As these technologies spread throughout American society more rapidly than copyright law could adjust, courts misclassified piano rolls and records as merely mechanics. Reducing these new fixations to machinery rather than embodiments of human expression left recorded human musical performances inadequately protected, which facilitated piracy.

Applying these lessons, this Article contends that generative AI training implicates copyright law in ways not dissimilar to the struggles copyright law encountered with the player piano and the phonograph. In order to avoid similar missteps, we argue that Congress, the federal courts, and the U.S. Copyright Office should act proactively and pragmatically to enable the law of copyright to regulate the unauthorized use of proprietary materials by AI developers. AI training sometimes feeds off data misappropriated from human creative expression with the aim of generating content that may reproduce the ingested works. This implicates the exclusive right of copyright holders to reproduction, as datasets assembled for training models copy protected works, raise risks of memorization, and can produce outputs that function as market substitutes.

We recommend that fair use should be assessed for purpose, amount, and market impact, while processes like voluntary and collective licensing should be given careful consideration as methods to ensure that AI can advance without misusing the creative works it relies upon.