By: Ana-Alicia Feng

introduction

Generative AI has taken the world by storm, for better or for worse. While not the first crossover of technology and the arts, generative AI’s current novelty and inexplicable and vast capabilities bring us once again to question how technology affects the authenticity of art and who can participate. Artists have always been at the forefront of either incorporating new technology into their art or at the very least commenting on how they see technology affecting them personally and on a larger societal level. Generative AI in an art context raises a multitude of legal questions, one of the more important ones being, who owns the art? Is it the AI-tech company? The person who inputs the prompts? Some mysterious third party? More specifically, with the U.S. Copyright Office declaring that AI cannot be credited as an author on a piece of art, how does this change the legal landscape for the AI-curious and artists who already use AI as part of their creative process?

understanding copyright

To fully understand the scope of the Office’s ruling, let’s review just what exactly it means to have copyright protection. Once a work of art is produced and fixed in tangible form, it is automatically protected by copyright law without any further action from the artist. Copyright deems that the owner (the creator) is within their legal right to recreate the work, sell and distribute copies, make new work from the original, and publicly display and/or perform the work, if applicable, among other things (Gormley 2020).

By choosing to register their work with the U.S. Copyright Office, the artist receives a certificate of registration and a public record of ownership. This documentation is particularly important, for example, in a lawsuit for infringement on an artist’s work (within the U.S.). Without copyright protection, AI-generated art is in the public domain and does not require permission of the “owner” to be reproduced, sold, etc. Importantly, copyright is only applicable to human creators, not monkeys, and not artificial intelligence (Gormley 2020).

how is a.i. complicating things?

In September of 2022, the U.S. Copyright Office accepted Kristina Kashtanova’s copyright registration for the comic book, Zarya of the Dawn. This was the first instance of registration for a comic book with images generated with a text-to-image AI program (Analla 2023). The issue? The Copyright Office didn’t know.

Five months later in February 2023, the Copyright Office canceled the registration of Zarya of the Dawn for Kashtanova’s failure to disclose the extent of AI’s contribution at the time of the initial registration approval. From the Office’s letter to Kashtanova’s attorneys, the Copyright Office states:

“We conclude that Ms. Kashtanova is the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements. That authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship. Because the current registration for the Work does not disclaim its Midjourney-generated content, we intend to cancel the original certificate issued to Ms. Kashtanova and issue a new one covering only the expressive material that she created.”

(Kasunic and U.S. Copyright Office 2023, 1)

Within that same letter, the Office referred to images from Kashtanova’s attorney’s initial letter to the Office to rebut claims that the Midjourney-produced images were altered by Kashtanova. In the example above, Kashtanova asserted that Photoshop was used to further adjust the image (specifically the lip). The Office noted that the alterations to the Midjourney image were too subtle to determine whether the image has sufficient human authorship. Overall, they concluded that the text, selection, coordination, and arrangement of the text created by Kashtanova can be registered for copyright as an unpublished literary work, as these elements exclude the work generated by artificial intelligence (Kasunic and U.S. Copyright Office 2023, 1-12). In other words, the comic book as a whole can be copyrighted, but the images cannot.

A second case further demonstrates the Office’s basis in human authorship. In 2018, Stephen Thaler, applied for copyright protection of the work, A Rare Entrance into Paradise, naming the AI program Creativity Machine as the sole creator. The Office denied the application on the grounds that copyright protection is granted for human authors, and on Friday, August 18, 2023, a U.S. federal judge upheld the Office’s ruling and rejected Thaler’s attempt to copyright the work (Naughton 2023).

While many generative AI programs state that users are given full rights to commercialize the images they create using the program, this does not constitute copyright. It simply means that users can proceed without fear of receiving a cease-and-desist letter and being prosecuted by the company for reprinting, selling, and merchandising the generated content (Mathur 2022). Full copyright protection must be sought with the Office.

In March 2023, the Copyright Office released additional clarification for application procedures regrading works that have elements of AI. The guidelines state that individuals applying for works with AI content can in fact claim copyright protection for their own contributions to the work. For example, a visual artist can claim sole authorship but must add a disclaimer that the selection, coordination, and arrangement of x-amount of content is of the artist themselves, and that y-amount of the work was generated by artificial intelligence (U.S. Copyright Office 2023).

This also means that any works that include AI-generated content that are pending approval and/or have already been approved, should be edited to reflect the correct disclaimer about authorship. Failure to accurately disclose AI contributions could result in a loss of copyright registration and legal protections (Belanger 2023).

The Photography Argument

Copyright for AI has been frequently compared to that of photography. The Supreme Court addressed photographic copyright in 1884, ruling that although photographs are created through a mechanical process, the image reflects the creative choices of the photographer and therefore deserves copyright protection (Lee 2023). A photographer may not be able to choose the angle of the sun, but they are in control of when they capture a moment and any alterations made to the photograph in the editing process. Unlike photography, however, AI-generated content establishes too much distance from the creator and has elements of unpredictability that establish the AI system as the entity making creative choices.

Who exactly does this affect?

Sougwen Chung builds and uses AI robots, dubbed D.O.U.G., to paint together on large canvases to explore human-machine partnerships. How is Chung’s work affected by the copyright ruling? Let’s take into account that the AI program and D.O.U.G. robots are of Chung’s design and are most likely eligible for patent, as compared to generative AI platforms for public use such as Midjourney and DALL·E. It’s also important to note that following the Office’s guidelines, the finished paintings are significantly manipulated by Chung herself and therefore most likely fall under eligible copyright protection. The only addition to a regular copyright application would be that Chung would have to disclose that a certain percentage of the work was created by artificial intelligence (Kaufman 2020).

It might, however, affect artists like Santiago Bautista González, an illustrator living in Guadalajara Mexico who offers AI commissions on Fiverr in addition to his hand-drawn orders. Given that Gonzalez’s personal touch is less identifiable, he would have to demonstrate his process in the copyright application. If Gonzalez is unable to copyright his work, then technically anyone on Fiverr (or anyone in general) can take the images of his work and sell them as their own, with no legal repercussion (Deck 2023).

In September 2023, 191 artists signed an open letter to Congress demanding that their voices be heard in the conversation around AI policy. The artists make the claim that generative AI lowers financial, physical, and social accessibility to art creation and allows artists to pioneer entirely new mediums. They also condemn the exploitative ways that generative AI can be used in the theft of the work of other artists and individuals and state that their interests are grounded in ingenuity and the creation of new art. It was not immediately clear how they intended to ensure that their AI-aided creative process would keep true to those values.

How does this affect the arts sector?

AI art has infiltrated the market and can no longer be extracted. Arts institutions now face the decisions to accept artwork that has input from AI. Given the nuanced and not entirely transparent decision-making process from the U.S. Copyright Office, it remains to be seen what standards will be used to determine how much of the “human touch” is required for art with AI content to receive legal protection.

Artwork without copyright protections may be more of a liability than an asset within a museum collection. The Getty has banned AI content for fear of the potential legal risk to customers (and in turn their own legal risk, naturally). In September 2022 the Colorado State Fair awarded a prize and then rescinded that same prize for Jason M. Allen’s submission of Théâtre D’opéra Spatial, a digital painting created with Midjourney (Roose 2022). Mr. Allen has refused to accept the decision. While the issue may remain unsolved for now, the Colorado State Fair and other similar organizations will have to consider clear policies and guidelines to avoid this kind of press and dilemma for future submissions.

In the public funding sector, grantmakers are already struggling with how to define art and measure the credibility of individual artists, and how much credibility should even be valued. The decision of whether to include artists who use generative AI has the potential to increase barriers to accessibility, but like the Colorado State Fair example, these decisions and their consequences will have to be a consideration for applicant eligibility moving forward.

Some issues that have been floating around include the aforementioned murky formula that the Office is using to determine sufficient authorship, but also the definition of artificial intelligence. Many commonly used software programs, such as Photoshop, are incorporating features that could be categorized as AI. The current ruling also incentivizes deception in the application process, which could cast general suspicion over all digital artwork regardless of whether it has any AI-generated content. It will also most likely discourage artists from using AI in their creative process and potentially stunt the progress of exploration of a new digital technique.

It is likely that the Copyright Office will have to continue amending their guidance, resources, and ruling, especially as AI-tech companies and pro-AI artists continue to demand a seat at the table. It’s also worth mentioning that many examples of avant-garde (non-AI) art incorporate technology and have large elements of randomization, accident or unpredictability, and distance from the artist.

If you are an artist or arts manager and want your voice heard, the U.S. Copyright Office has opened its public comment period for opinions on issues regarding AI and intellectual property. They are accepting email and paper mail from August 30, 2023 until the end of 2023.