A Trio of Decisions on the Copyrightability of Ai-drawn Art Should Prompt Artists to Stress Human Involvement if They Hope to Generate a Copyright

• Publication year: 2024
• Citation: Vol. 7 No. 3

[Page 195]

T. Chase Ogletree*

In this article, the author discusses three decisions that illustrate the difficulties an artist will face when trying to copyright an image created by a computer program.

A Tale of Two Artists

Scenario One: Jane is an artist who wants to create a computer program that generates images based on her text prompts. So, she follows a series of steps. Step 1: She selects a data set of existing artwork. Step 2: She feeds those images to her program's machine learning algorithm so it can learn the art style and patterns. Step 3: She enters a few test prompts to see what images are generated. She then repeats Steps 1 and 2, even tweaking the algorithm itself, until the generated images begin to approximate her prompts. Step 4: She tweaks her prompts dozens of times to eventually get the precise image she wants. Step 5: She makes minor revisions to the image in an image-editing program.

Scenario Two: Frank, who is also an artist, wakes up and groggily looks down at his comforter, the lumps of which remind him of mountains. Frank reaches for his smartphone, takes a picture of the "mountains," then goes back to bed.

Which artist is more likely to get their image copyrighted? The answer is undoubtedly Frank. The U.S. Supreme Court ruled that photographs are copyrightable more than a century ago. By contrast and even though Jane clearly did more work to create her image, a nonhuman computer program generated her art, which

[Page 196]

means she would need to show that she was sufficiently involved in the process, including that she "dictated," not just "influenced," the final image. And the evidence required to make such a showing is, at this point, still uncertain.

Three 2023 decisions have illuminated the obstacles Jane faces in obtaining a copyright, as well as a potential path forward.

Thaler v. Perlmutter ¹ Holds That 100 Percent AI-Generated Art Is Not Copyrightable

Plaintiff Stephen Thaler owned a computer system called the "Creativity Machine," which he claimed generated a piece of visual art on its own. Thaler sought to register the work for copyright, listing the computer system as the author and explaining that the copyright should transfer to him as the machine's owner. The Copyright Office denied the application on the grounds that the work lacked human authorship. Thaler then challenged the decision in federal district court under the Administrative Procedure Act.

The court upheld the U.S. Copyright Office's decision, finding that human authorship is an essential part of a valid copyright claim. While the court acknowledged that copyright was supposed to adapt with the times and that expression generated by new forms of technology, such as photography, had received protection in the past, it ultimately concluded that copyright had never stretched so far as to protect works generated by new forms of technology operating absent any guiding human hand. Rather, for any new type of work to fall within the bounds of copyright, a human must be involved in, and have ultimate creative control over, the work. The court further reasoned that copyright's main purpose—incentivizing individuals to create and invent—does not apply to nonhuman actors as they do not respond to such incentives. It also rejected Thaler's argument that he should own the copyright under the work-for-hire doctrine or common law property principles because he owned the Creativity Machine, finding there was no valid copyright to transfer in the first place.

In addition, Thaler noted its holding was in line with other court decisions denying copyright to nonhuman entities. For example, in 2018, the Ninth Circuit found that only humans had standing to pursue copyright claims in the famous "Monkey Selfie" case. In 2011, the Seventh Circuit refused to recognize a copyright in a

[Page 197]

cultivated garden. And multiple courts have found that celestial beings were not entitled to copyright protection. Notably, in some of the celestial-being cases, the courts ultimately granted the alleged celestial works copyright protection because there was sufficient human involvement, such as (1) formulating the specific questions asked to the spirits and then selecting and arranging the "revelations," or (2) working with human collaborators to revise and edit notes from a conversation with a celestial voice. This leads to the next crucial point in discussing the copyrightability of Al-drawn art.

Thaler Leaves the Door Open for Copyright Protection in AI-Drawn Art if the Applicant Can Show Some Degree of Human Involvement

The court conceded that the rise of AI "will prompt challenging questions regarding how much human input is necessary to qualify the user of an AI system as an 'author' of a generated work." ² The court then explained, however, that Thaler's case was not nearly so complex. While the plaintiff asserted new facts—that (1) he provided instructions and directed the AI to create his work, (2) the AI was entirely controlled by him, and (3) the AI only operated at his direction—Thaler did not present those human-involvement facts to the Copyright Office, and thus the court could not consider them. By contrast, Thaler had represented to the Copyright Office that the work was created autonomously by a machine, and that the plaintiff's copyright claim was solely based on his ownership of the machine. Based on those facts, the court ultimately held that in the absence of any human involvement in the work's creation, the plaintiff's AI-generated work cannot be copyrighted. And while Thaler is now appealing the court's decision, his success is highly questionable based on the facts he originally presented.

But this holding implies that a work could be copyrighted if there was some degree of human involvement in the work's creation. This is especially true considering some celestial-being cases granted copyright protection to works that were at least partially the product of human creativity. But the devil is in the details. Where do you draw the line to determine how much human input is needed for a copyright? Fortunately, the Copyright Office has

[Page 198]

recently issued two decisions (and other guidance) to help answer that question.

The Copyright Office Rescinds a Copyright for AI-Drawn Comic Book Art ³

Kristina Kashtanova submitted a comic book called Zarya of the Dawn (the "Work") to the U.S. Copyright Office. It was registered the same day. Shortly after, the Copyright Office became aware of Kashtanova's social media statements that she created the comic book art using the AI program Midjourney. Because her original application had not explicitly disclosed the use of artificial intelligence (AI), the Office notified Kashtanova that her registration would be cancelled unless she provided additional information.

Kashtanova explained her role in creating the comic book and argued that (1) she authored every aspect...