•  
  •  
 
Emory Law Journal

Abstract

Since it began to pick up steam in the mid-2010s, generative Artificial Intelligence (AI) has quickly developed breathtaking capabilities. It has found applications in a great variety of industries, and in the last couple of years it has developed to the point where it has captured public attention by producing artistic works that match the quality of human-made works. While bringing great advantages, this development is not without concern, especially within the artistic community. Much has been made of the use of artists’ works in training generative AI, the likely replacement of artists with AI models, and the threat these pose to artists’ livelihoods. Unfortunately for artists, copyright law provides practically no restrictions on the use of their works in training AI or on the replication of their artistic styles in AI-generated works.

This Comment proposes using copyright law to address some of these concerns, without significantly inhibiting the progress brought by AI. After first providing a background of copyright law and generative AI, this Comment surveys the inadequate theories of infringement put forth by plaintiff-artists in recent AI-related copyright litigation. It then gives a much-needed definition to artistic style and shows that style can, and should, be protected under existing copyright law. Finally, this Comment proposes a three-step test for style infringement that follows the established copyright infringement test. The proposed test, like the established test, begins with copying-in-fact and ends with substantial similarity. An intermediate step ensures that the protection of style does not overreach and interfere with AI-produced works that do not specifically attempt to replicate an artist’s style. This test borrows theories from character copyright to slightly expand copyright law and protect artistic style—a protection that is sorely needed in the rapidly evolving area of generative AI.

Download

Included in

Law Commons

Share

COinS