As AI systems demonstrate unprecedented capabilities to create, manipulate, and generate original content, the interplay between AI and copyright law has come to the forefront of legal discourse. This convergence presents both challenges and opportunities as stakeholders grapple with defining authorship, ownership, and infringement in a landscape where machines are capable of producing works that rival human creativity. As the technology continues to advance, sprouting legal questions along the way, it is important to establish a foundational understanding of the current state of affairs at the intersection of AI and copyright law. This article will address one of the many AI-related questions – the extent to which humans can own the copyrights in content that AI creates or helps to create.
To best understand the answer to this question, it’s useful to review the basics of copyright law itself. The basis for copyright law in the U.S. comes from the Constitution, specifically Article 1, Section 8, Clause 8 which states that it is among the powers of Congress “[t]o 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.” Congress codified earlier judicial interpretations of this language into the US Copyright Act, which itself has been subsequently interpreted by later courts. As a result, there are a few requirements for establishing a copyright in the U.S. First, authorship, which includes the related requirements of originality and minimal creativity. Second, fixation, which requires that a copyrightable work be fixed in a tangible medium of expression.
The latter requirement has already undergone modern interpretation in light of technological advances. Namely, the question arose of how to treat digital works that are not “tangible” in a traditional sense. This question was answered in Williams Electronics v. Artic International, which held that works need only be “‘sufficiently permanent or stable to permit it to be … reproduced, or otherwise communicated’ for more than a transitory period.” This holding originally related to video games, but applies for all varieties of digital works, including digital AI works.
The first requirement, authorship, has been under much more scrutiny in recent months. Notably, the question of who can be an author has been a hot topic in copyright law since 2017 when the United States District Court for the Northern District of California ruled that animals could not be authors for the purposes of the Copyright Act. This ruling was subsequently upheld by the Ninth Circuit on appeal. While non-human animals cannot be authors, the question remained whether other non-human entities likewise lacked the capacity for authorship. In 2019 a computer scientist named Stephen Thaler attempted to register an AI-generated work titled “A Recent Entrance to Paradise.” An AI called the “Creativity Machine” allegedly created the visual work entirely autonomously, and so Thaler listed the AI as the author. The Copyright Office rejected the application on the grounds that the Creativity Machine lacked necessary human authorship. In August of 2023, the United States District Court for the District of Columbia upheld the office’s rejection, establishing a judicial precedent against AI as authors. The decision reflects a unified approach to the question of AI authorship between the courts and the Copyright Office which continues to underscore the necessity of human creativity.
Besides rejecting Thaler’s application, the Copyright Office previously canceled and then reissued a modified registration in the case of a comic book author who claimed that an AI had generated the artwork used in her comic. After learning that the images in Kristina Kashtanova’s comic, Zarya of the Dawn, were AI-generated, the office rescinded its initial registration of the entire work and issued a more limited copyright registration covering only the “expressive material” of the human author, such as the text, the selection, and arrangement of the images. So, while AI is ineligible to create or possess copyrights in the work it generates, this does not mean that any works contributed to by an AI are entirely ineligible for copyright protection. Human authors who contribute creatively to a work partially generated by AI still engender copyrights in their contributions. This paradigm supports the enterprise of human creativity, at least as defined by the Copyright Office, but does more to distinguish AI from traditional tools for creators. AI algorithms are more than mere paintbrushes or musical instruments that a human author uses to express their creativity. When AI is used in the creation of works, it is not entitled to authorship itself but, it may serve to remove some aspects of authorship from the human author who wields it.
When creating works using AI, the user is well-advised to carefully delineate and document the human contribution to the work. That contribution may be what is traditionally considered creating, such as writing the prompt. Or the user’s contribution may be creative in the sense that the user selects which portion of the AI content to use and chooses to present it in a certain order. In the comic book example, AI generated the images, but the human author chose which images to use and the order to put them in and the human author wrote the text to accompany the images. In a pure text example, the user may write the prompts and then heavily edit the AI generated content. The user likely has copyright protection in their input and even in the final content because of the extensive editing. It seems analogous to a co-author situation which results in each author having copyrights.
As the Courts grapple with this new technology, the extent of copyright protection granted to a human user of AI will continue to be clarified. For now, we apply old law to new technology to anticipate new law.
Co-Author: Jackson Polansky, a research and writing consultant specializing in intellectual property.