The current boom in artificial intelligence (AI), including programs like DALL-E and ChatGPT, has sparked widespread conversation and debate about whether and when AI-generated works (or AI “output”): (1) should be copyrightable; and (2) infringe upon the works (or data) used to train the AI.
A recent decision by the U.S. Copyright Office (available here) shed some new light on the first question. Here’s the backstory:
- On September 15, 2022, artist Kris Kashtanova received a U.S. copyright registration for her comic book, Zarya of the Dawn, which includes images generated by Midjourney, an AI program that turns text input or “prompts” into digital works of art. In her copyright application, Kashtanova listed herself as the author of the comic book and did not mention that she used AI to create any part of it.
- The Copyright Office later became aware of Kashtanova’s social media posts explaining how she used Midjourney to help create Zarya. On October 28, 2022, the Copyright Office notified her that it intended to cancel the registration because her application was incorrect or substantively incomplete.
- Kastanova responded on November 21, 2022 with a letter detailing her use of Midjourney in connection with the comic book, arguing that the registration should not be canceled because she merely used Midjourney as an “assistive tool.” In the alternative, she argued that portions of Zarya are copyrightable due to her creative selection, coordination, and arrangement of elements.
On February 21, 2023, the Copyright Office revealed it will cancel Kashtanova’s previous registration for Zarya due to her “failure to exclude non-human authorship contained in the work,” and replace it with a new registration covering only the original authorship she contributed: specifically, both her original text and the selection, coordination, and arrangement of that text with the AI-generated images.
The new registration will explicitly exclude “artwork generated by artificial intelligence.” But why, exactly? Despite Kashtanova providing “hundreds or thousands of descriptive prompts” to Midjourney to generate “as perfect a rendition of her vision as possible,” according to the Copyright Office, Kashtanova was not the “author” of the resulting images for copyright purposes. This is because, “unlike other tools used by artists” (such as Adobe Photoshop), Midjourney generates images using prompts in an “unpredictable way.” As the Copyright Office put it, “[b]ecause of the significant distance between what a user may direct Midjourney to create and the visual material Midjourney actually produces," Kashtanova did not have enough control over the final images generated to be the “inventive or master mind” behind them.
Though the Copyright Office did not take a position on the protectability of the prompts themselves, it rejected the argument that use of human-authored text prompts allows for copyright protection of the resulting images. By analogy, the Copyright Office noted that if Kashtanova had hired a visual artist to create a character and provided the artist with precisely the same prompts that she used in Midjourney, Kashtanova would not be the author of the resulting image (unless the image qualified as a work made for hire); instead, “the author would be the visual artist who received those instructions and determined how best to express them.” Nothing was different, in the Copyright Office’s view, when Kashtanova used Midjourney.
Importantly, the Copyright Office clarified that its decision was based on the specific way in which Midjourney operates, and that creators using “other AI offerings” may see a different result and become the author of AI-generated images for copyright purposes. Creators using AI in any capacity to generate content—including images, videos, music, code, scripts, books, games, and beyond—should endeavor to understand how the specific technology works and what rights they may or may not have in the final product.
Creators using AI in any capacity to generate content—including images, videos, music, code, scripts, books, games, and beyond—should endeavor to understand how the specific technology works and what rights they may or may not have in the final product.