Authorship and Copyright In Hybrid AI-Human Collaborative Works

Kashtanova wrote the text of the work but used the Midjourney artificial intelligence (AI) technology to generate the work's constituent images. Midjourney, like other "generative" AI technologies for creating images, was preliminarily trained to generate its output images by analyzing an existing corpus of visual works. To generate images using the tool, users feed Midjourney a textual prompt representing desired concepts or qualities of the output image (e.g., "snow mountain wolf growling, face view, snow fog mist, hyper-realistic") and may also provide a URL of an image and/or parameters defining desired aspect ratio or other properties. Midjourney generates four images based upon the user-supplied prompt, any one of which the user can then iterate upon further with new prompts. The final output images evolve from initial "noise" values that are based on a randomly generated seed, so the generative process is by default non-deterministic. However, the process can be made deterministic if the user explicitly specifies the seed value. Kashtanova created each of the images in the comic book through an iterative process of first supplying prompts to Midjourney that described the desired content of the output image, then selecting one of the four images output by Midjourney, and then revising her prior prompt to refine the image in the next iteration, until eventually arriving at a finished image.

The Copyright Office noted that its earlier ruling regarding the AI-generated artwork A Recent Entrance to Paradise had already established that images generated entirely by generative AI algorithms, without the assistance of human input, lack a human author and are thus not copyrightable under the 1976 Copyright Act. Kashtanova's work, in contrast, presented arguably more complicated legal issues due to Kashtanova's involvement, particularly in her direction over Midjourney's generation of the images through issuance of iterative prompts. The Office concluded that although the text accompanying the images in Zarya of the Dawn, as well as the selection and arrangement of the images, contained sufficient creativity to be afforded copyright protection, the individual images did not. The ruling focuses on authorial control and image predictability, citing the "master mind" concept from Burrow-Giles Lithographic Co. v. Sarony, the Supreme Court decision which in 1884 extended copyright protection to photography, and argues that the link between Kashtanova's textual prompt and the ultimate AI-produced output images is too tenuous for the user to be the master mind behind those images.