Copyright Office Draws a Line in Shifting Sand on AI

Part 2 of its 3-part report addresses copyrightability of AI-generated work

Anyone who has followed the U.S. Copyright Office’s now two-year long proceeding on copyright and artificial intelligence will not be greatly surprised by the Office’s conclusions regarding the copyrightability of works produced by or with the aid of AI. As Cornell Law professor James Grimmelmann pithily summed up the main takeaway from last week’s release of Part 2 of the planned 3-part report, “If you make art with the help of AI, it’s copyrightable. If you ask AI to make art for you, it’s not.”

Less pithily, the 42-page report concludes that the use of AI-powered tools by a human artist or author to assist in realizing or enhancing her own original creative expression, by itself, does not disqualify the overall work from copyright protection, although any AI-generated elements may not separately be protectable. Nor should the use of assistive AI tools prevent the overall work for being licensed for use by third parties. At the same time, works produced by entirely by AI, or with de minimis human involvement, are not eligible for copyright protection.

The findings are generally good news for the Hollywood studios, which had expressed concerns in comments to the Copyright Office that the increasing use of AI tools in production and post-production might render large swaths of expensively produced films ineligible for protection. The report even specifically cites by way of example, that “a film that includes AI-generated special effects or background artwork is copyrightable, even if the AI effects and artwork separately are not.”

The report holds less good news for those who argued the crafting, revising and entering of iterative prompts into an AI model until the desired result is achieved should qualify as authorship, and the final output therefore should be protectable. “The Office concludes that, given current generally available technology, prompts alone do not provide sufficient human control to make users of an AI system the authors of the output,” the report counters. “Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output.”

The Office has been telegraphing its latest position since issuing its initial guidance for registering works containing material generated by AI in March 2023, which first declared its policy of excluding from protection those elements of a work produced without direct human control. It reaffirmed that view in defending its ruling to deny registration to Stephen Thaler’s image “A Recent Entrance to Paradise,” which he claimed was “autonomously created” by a generative AI system, and again in partially rescinding its previous registration of the graphic novel Zarya of the Dawn over the author’s use of Midjourney to create the illustrations.

But as was true then is true now: there are devils lurking in the details. The most obvious area for mischief concerns just how much human input in required, and of what type, to render an AI-assisted work copyrightable. The Office essentially punts the question.

The report leans heavily on the inherent unpredictability to the workings of AI’s “black box” algorithms to justify the ambiguity. “The fact that identical prompts can generate multiple different outputs further indicates a lack of human control. As one popular system explains on its website, ‘[n]o matter how detailed . . . the same text describes an infinite number of possible” outputs,’” the report explains. “Repeatedly revising prompts does not change this analysis or provide a sufficient basis for claiming copyright in the output…[I]nputting a revised prompt does not appear to be materially different in operation from inputting a single prompt. By revising and submitting prompts multiple times, the user is “re-rolling” the dice, causing the system to generate more outputs from which to select, but not altering the degree of control over the process.”

So what would demonstrate a sufficient degree of control? “[W]here a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author of at least that portion of the output. Their own creative expression will be protected by copyright, with a scope analogous to that in a derivative work.”

That basically just restates the question. How much perceptibility is needed, how and by whom? It’s a formula for a lot of administrative hair-splitting, and a likely invitation to litigation over where individual hairs get split. As the Office rightly points out, insofar as all applications for registration are reviewed before registration is granted, a certain amount of line-drawing is par for the course. But that line will only get fuzzier as the technology gets more powerful and the role of humans in the creative process evolves.

Nor should creators and rights holders look to Congress for greater clarity on the issue, according to the report. “Even if Congress were to consider addressing this issue through legislation, greater clarity would be difficult to achieve,” the report reads. “Because the copyrightability inquiry requires analysis of each work and the context of its creation, statutory language would be limited in its ability to provide brighter lines.”

Oh.

In the end, the Office falls back on the courts to bring some clarity to its own policy. “[T]he courts will provide further guidance on the human authorship requirement as it applies to specific uses of AI (including in reviewing the Office’s registration decisions),” according to the report.

The courts could be forgiven for hoping the copyright experts at the Copyright Office would provide further guidance as to how to analyze the question.

To its credit, the Office stipulates at points throughout the report that its current conclusions reflect the current state of the art in generative AI technology. It acknowledges that future developments could necessitate reevaluation of some or all of its views.

The report also claims the Office will introduce revised guidance on registration later this year based on the findings from its two-year proceeding, along with an update to its Compendium of U.S. Copyright Office Practices. As things stand now, however, the increased use of AI tools in creative processes will likely come with an increased urgency to reengineer internal creative workflows, record-keeping systems, financial models, royalty and residual payment systems, and licensing practices within media companies and industries, and among individual creators, to comply with the policy of treating human- and AI-produced elements differently — even within the same work — for copyright purposes.

On that, at least, there’s a measure of clarity.