Artificially Yours: Who Owns Rights in AI-Generated Art?

By  on 

Don’t worry, machines haven’t completely replaced humans as artists—at least, not yet. But the U.S. Copyright Office is considering the possibility.

The Copyright Office recently declared that it will not grant protection over AI-generated works, upholding its longstanding rule that non-human authors cannot own copyright. At the same time, the Office is well aware that AI technology is changing everything. It is seeking public input on AI-related issues throughout the month of May, starting with a listening session on AI and visual arts from 1 to 4 pm Eastern today. A session on audiovisual works is scheduled for May 17, and another for music and sound recordings on May 31. Members of the public can access Zoom links through the Office’s website.

As it stands, under U.S. law there must be “some element of human creativity” for material to be copyrightable. In a registration decision issued February 21, 2023 regarding a comic book entitled “Zarya of the Dawn,” the Copyright Office explained that it would only issue a registration covering the compilation of images and text created by the human author. The registration would not cover AI-generated images contained within the compilation.

The decision relied on the 1884 U.S. Supreme Court case Burrow-Giles Lithographic Co. v. Sarony, in which the Court grappled with what was then the newly emerging technology in art: photography. Although the process of the camera capturing the photograph was purely mechanical, the Court explained, the photographer’s artistic choices in composing the photograph were protectable because they were instrumental in the ultimate output. The Copyright Office drew a parallel to Zarya of the Dawn, explaining that “the Work is the product of creative choices with respect to the selection of the images that make up the Work and the placement and arrangement of the images and text on each of the Work’s pages.” In contrast, the images themselves were created using Midjourney, an AI tool that generated visual depictions based on text prompts. “While additional prompts applied to one of these initial images can influence the subsequent images, the process is not controlled by the user because it is not possible to predict what Midjourney will create ahead of time,” the Office concluded. Therefore, the process of creating the images “is not the same as that of a human artist, writer, or photographer.”

The Office followed up with formal guidance published in the Federal Register on March 16, 2023, explaining that the term “author” as used in the Intellectual Property Clause of the U.S. Constitution and in the Copyright Act “excludes non-humans.” Therefore, when applying for a copyright registration of a work that contains AI-generated content, the AI-generated content must be explicitly excluded from the application. If the applicant fails to disclose which portions of the work were AI-generated, and the Office later becomes aware of that fact, the resulting registration may be cancelled.

In fact, this is what happened in connection with Zarya of the Dawn. After originally obtaining a registration covering the entire comic book, the book’s human author boasted about it on social media, generating interest among journalists. The Copyright Office saw the attention its original decision had gotten, and decided to revisit the issue sua sponte, revoking part of the registration.

The Zarya decision does provide helpful guidance on the current state of copyright law and AI, but the lines between human-generated art and AI-generated art remain blurry. For example, although Midjourney actually created the output of images in Zarya, a human had to come up with the text prompt to initiate the process. Should the human author be allowed to claim copyright in the prompt? And, if so, would that right actually be meaningful?

Generally, it is not possible to obtain copyright in short phrases. The Copyright Office has concluded that short phrases do not contain a sufficient amount of creativity to be copyrightable. Even so, depending upon the length and complexity of a given prompt in an AI tool, there may be room to argue that authors are making creative choices affecting the ultimate output generated by the tool. If a particular tool generates consistent output based on the same prompt, copyright could help artists exclude others from using identical prompts to obtain identical outputs.

At the same time, under the Copyright Office’s current policy, there would be no copyright in the output itself. In other words, no one would own rights in that output, so copying it wouldn’t constitute infringement. Therefore, even if the artist were able to prevent another individual from using a particular prompt to re-generate an output, there would be nothing stopping that same person from simply making a digital copy of the output itself. Copyright in the original prompt would be toothless in this context.

It will be interesting to see how all of these dynamics play out as artists and authors increasingly rely on AI tools as part of their creative process. It will also be interesting to see how the Copyright Office balances the current legal framework preventing copyright over AI-generated content with the original goal of copyright law as articulated in the Constitution: “To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.”