Award-winning AI Art Fails to Get Copyright Protection on a Technicality

An AI-generated artwork that won a digital art contest at last year’s Colorado State Fair and went viral has been denied copyright protection in the U.S.

The Copyright Office denied protection to Jason M. Allen, who created the award-winning work ‘Théâtre d’Opéra Spatial' using Midjourney, on the grounds that it “lacks human authorship.”

The Copyright Office ruled that Allen was not the author of the Midjourney image “because his sole contribution to the Midjourney image was inputting the text prompt that produced it.”

Allen tried to argue that he made multiple revisions of the image – and that edits he made to the image using Adobe Photoshop were enough to show “original authorship.”

He also argued that the Copyright Office placed “a value judgment on the utility of various tools” when creating and editing works.

The Copyright Office, however, rejected his arguments, adding that Allen failed to disclaim the material produced by Midjourney – applications for copyright protection are now required to disclose if a work contains “more than a de minimis amount” of AI-generated content.

“Because the authorship in the Midjourney Image is more than de minimis, Allen must exclude it from his claim. Because [Allen] has refused to limit his claim to exclude its non-human authorship elements, the Office cannot register the work as submitted,” a ruling read.

The ruling comes as Allen is attempting to capitalize on the image’s notoriety by selling limited edition versions of the image on his website priced at $500 each.

AI and intellectual property concerns growing

The Copyright Office’s decision is its latest denial of protection for an AI-generated work. The rise of AI image generation tools has seen the agency inundated with applications for copyright protection.

Notable refusals include a denial for Stephen Thaler’s Creativity Machine, which was later affirmed in court, and the case of Zarya of the Dawn, a comic book that was also created with the help of Midjourney.

The Copyright Office has since updated its examination practices regarding the level at which AI tools were used to create the final version of a work. The clarification means that if a human applicant disclosed the use of AI tools and their use did not outweigh the human input, it may be granted U.S. copyright protection.

Intellectual property is fast becoming one of the biggest concerns in AI. In a survey published this week, GitLab found that IP was among the biggest concerns for companies adopting AI.

The concerns are around either protecting existing IP – keeping trade secrets from becoming public after inputting them into a language model − or using an AI model that was trained on copyrighted material without express permission from the author.

Just last month, it emerged that the Books3 dataset used to train prominent AI models, including Meta’s LLaMA, GPT-J and Bloomberg’s BloombergGPT, illicitly contained works from authors such as Stephen King and Margaret Atwood.