AI, Copyright Law and the Requirement of Human Authorship

The U.S. Copyright Office recently proclaimed that most forms of AI-generated content are not entitled to copyright protection. This latest directive came in an official policy titled “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence,” which effectively excludes AI-generated works from receiving copyright protection because they lack a human author.

This categorical exclusion reaffirms the long-standing requirement of human authorship that gained national attention in 2014, when a crested macaque monkey took a selfie using wildlife photographer David Slater’s camera and sparked a debate over who owned the copyright.

Fast forward to 2023, the most talked-about topic in the legal community is about AI. Its unprecedented pace of development, combined with an array of previously unresolved legal questions, has led experts in intellectual property law to consider the implications for creators using AI and generative AI systems.

Generative AI is a type of AI model that allows users to create content such as digital artwork, text, audio, video, speeches, poems, short stories and other media using chatbots. Common, and increasingly familiar, examples include ChatGPT, Bing Chat, and Bard, as well as AI art systems such as Stable Diffusion, Midjourney, and DALL-E. Writers have used images generated by AI to create graphic novels, and others have taken to publish AI-written e-books on Amazon.

Yet, creators using AI must deal with the reality that the vast majority of AI-generated content is not protected by copyright, and creators who incorporate it into their works may not have the ability to prevent unlicensed or unaccredited usage of the content.

Requirement of human authorship

Copyright is a form of intellectual property that protects “original works of authorship” from the moment the work is “fixed in any tangible medium of expression” and provides those who do so with the exclusive legal right to publish, perform, or record various works of expression, including text, photographs, video and audio recordings, artwork and software code.

Copyright law also permits original authors to sell their copyright and to grant licenses so that others can publish or otherwise use or reproduce the copyrighted material. Publishing a copyrighted work (for example by posting it on the internet, copying or distributing the work) without a license from the author constitutes copyright infringement.

Authors can register their works with the U.S. Copyright Office to gain additional legal rights and benefits, which include creating a public record of the work’s origin and a legal presumption ownership, and providing the named author with the right to bring a federal lawsuit for copyright infringement and recover statutory damages and attorney’s fees.

Copyright cases decided in the 19^th Century tended to focus exclusively on protecting “the fruits of intellectual labor” which were “founded in the creative powers of the mind.” However, the Copyright Act of 1976 explicitly recognized the vital role that technology plays in fixing or capturing copyrighted works, by describing mediums of expression as sources from which the work “can be perceived, reproduced or otherwise communicated, either directly or with the aid of a machine or device.”

Despite this statutory language, courts and the Copyright Office have continued to grapple with drawing the boundaries of copyright protection when technology is involved, refusing to grant copyright protection to a tie-dye process as well as artwork created by a computer algorithm, but granting protection to a photograph edited with Adobe Photoshop.

The question of whether copyright protection exists in the absence of a human author came to a head in 2011, when British wildlife photographer David Slater encountered a remarkably confident monkey named Naruto who used Slater’s camera to snap a selfie.

Stay updated. Subscribe to the AI Business newsletter

When Wikimedia Commons sought to publish the photograph for distribution in its collection of free online images, Slater claimed ownership, which led to a lawsuit between PETA, Slate and Wikimedia. PETA claimed in the lawsuit, which has become widely known as “the Monkey Selfie Case,” that the monkey Naruto was the rightful copyright owner. In a 2017 decision, U.S. District Judge William Orrick found “no indication” that the Copyright Act extended to animals and held that Naruto could not own a copyright.

The Copyright Office followed suit and declared that the monkey’s photograph was not eligible for federal copyright registration and that only works created by human beings — not those created by nature, animals or even the Holy Spirit, as alleged in a 1995 copyright case — could be registered. As a result of these rulings, works lacking a human author cannot be copyrighted and, therefore, can be freely used by anyone.

The Copyright Office’s guidance on protecting AI

Since the unfortunate Naruto lost his copyright battle, the Copyright Office has confirmed that works created “without any creative contribution from a human actor” cannot be registered for copyright protection.

For example, in 2018, the Copyright Office rejected an “application for a visual work that the applicant described as ‘autonomously created by a computer algorithm running on a machine.’” Most recently, in February 2023, the Copyright Office found that images in a comic book called Zarya of the Dawn, which were generated by the text-to-image engine Midjourney, could not be protected by copyright. However, the office found that the author was entitled to limited copyright protection on the text, which she had written, and on the selection, coordination and arrangement of the work, which she had fashioned to incorporate the AI-generated images.

Finally, in March 2023, the Copyright Office published a statement of policy which expressly excluded AI-generated content from copyright protection, save for a few exceptions regarding works that transform AI-generated content with sufficient creative expression.

The Copyright Office clarified its process of evaluating an application for works involving AI, which involves a determination of whether the program is merely “an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) (are) actually conceived and executed not by man but by a machine.”

The Copyright Office also stated that it will consider on a case-by-case basis whether the asserted “AI contributions are the result of ‘mechanical reproduction’ or instead of an author’s ‘own original mental conception to which (the author) gave visible form,’” depending on “how the AI tool operates and how it was used to create the final work.”

Indeed, in certain non-AI applications, depending on the degree of human involvement, even a work that is generated in part by a non-human author, such as a “celestial being” as alleged by a churchgoing-author in a 2003 copyright case, can gain protection.

Thus, given the possibility that works incorporating non-human input may be protected in some fashion by copyright, one may wonder about the level of protection available where a user prompts an AI system with a detailed and specific set of instructions, so as to convey a particular creative vision for the output. As applied to programs such as ChatGPT or Dall-E, which utilize textual prompts, the Copyright Office takes a limited view of human contributions to generative AI.

Using ChatGPT vs. Adobe Photoshop

Citing its understanding of the generative AI technologies currently available, the Copyright Office has determined that “users do not exercise ultimate creative control over how such systems interpret prompts and generate material” and that “when an AI technology determines the expressive elements of its output, the generated material is not the product of human authorship.” Therefore, prompt-based AI systems as currently configured simply do not generate copyrightable material because they, rather than the human operator, create the output.

In reaching this conclusion, the Copyright Office specifically distinguished how artists create works with assistance from other technology such as Adobe Photoshop. Users of Photoshop or other well-established tools select the visual material they wish to modify, choose which tools to use and what changes to make, and then take specific steps to control the final image such that the resulting output amounts to the artist’s “own original mental conception, to which (they) gave visible form.”

By contrast, even generative AI programs that require the user to input detailed prompts do not generate copyrightable material because, ultimately, it is the AI and not the human user who controls and generates the final output.

Given that the main goal of copyright protection is to reward creative expression, the difference between rewarding human creators versus those who query ChatGPT, Dall-E or similar AI programs comes down to recognizing what is and, just as importantly, what isn’t creative expression.

At least for now, human creators remain the only candidates who may obtain copyright protection and the law simply does not permit the delegation of creative acts to non-human AI actors.