The United States Copyright Office recently rejected Ankit Sahni's Second Request for reconsideration to register his AI-generated artwork, "Suryast." This is a unique case in the realm of AI-generated art, primarily due to the artwork being derived from Sahni’s own original photograph utilizing RAGHAV AI painting software.
RAGHAV AI painting software allows a user to create an image with the same content as a base image but in the style of a chosen picture. Sahni used the RAGHAV to transform Sahni's original photograph in the style of Vincent van Gogh’s “The Starry Night.” Sahni's inputs were limited to taking and selecting the photograph the image is based on, the style image (The Starry Night), and setting a value for the level of style transfer RAGHAV was to use. While the resulting image was a depiction of his photograph in The Starry Night style, Sahni's amount of human involvement raised questions about the threshold of creativity necessary for copyright protection.
The Copyright Office rejected Sahni's copyright claim for his AI-generated art, asserting non-human creations are ineligible for copyright. The Copyright Office stated that whether a work is sufficiently human-created is case specific, but if AI is responsible for “traditional elements of authorship,” the work is not copyrightable.
While Sahni argued that his work, including his original photo, should be considered as a whole, the Copyright Office disagreed. They utilized a derivative work analysis that “focuses on the new authorship contributed by the derivative author to that work — rather than the authorship from the preexisting work that may have been incorporated into the derivative work.”
Sahni's limited inputs to RAGHAV, including the original photo, "The Starry Night," and the style transfer level, led the board to determine that RAGHAV, not Sahni, dictated the artwork's outcome. Sahni argued that elements such as the sunset, clouds, contours of a building, and the sky (which accounts for the upper two-thirds of the work), as well as the style of Van Gogh, were under his creative control. The board found that Sahni lacked control over the final depiction, including where those elements would be placed, if they appeared at all, and what colors would be used, thus undermining claims of human authorship. Additionally, Sahni's choice in style transfer strength was deemed insufficient for copyright protection, and his selection of various inputs and settings were classified by the board as unprotectable ideas, not creative decisions.
While this decision was consistent with the board’s previous decisions, it is unique in that it provided further guidance on the level of human input required to obtain a copyright. The board also noted that the analysis is case-by-case and year to date, the Copyright Office has granted approximately 100 applications containing AI-generated material, where the AI-generated portions were disclaimed.
It will be interesting to see if applicants with greater human involvement in an AI system's creative process can achieve copyright status without the requirement of a disclaimer.
This article is part of a series of client advisories that discuss the potential benefits and drawbacks of generative AI in the workplace, relevant statutory and regulatory guidance, litigation risks, and other critical issues.
