Skip to main content
Top Button
Cyborg Authorship and Artistic Expression? Copyright Office Issues Guidance on AI-Generated Art Cyborg Authorship and Artistic Expression? Copyright Office Issues Guidance on AI-Generated Art

Cyborg Authorship and Artistic Expression? Copyright Office Issues Guidance on AI-Generated Art

In the latest chapter of the Copyright Office’s struggle to address the Constitutional and statutory subject matter controversy arising from works that are, at least in part, generated through artificial intelligence (AI), the Office issued a new Copyright Registration Guidance on Works Containing Material Generated by Artificial Intelligence on March 16, 2023 (the “Guidance”). This was prompted by the Copyright Office’s receipt of registration applications that either: (1) name AI technology as the author or co-author (or otherwise include AI attributions in the application), or (2) mention AI technologies in the title of the work or in the “acknowledgments” section of the deposit.

Generative AI describes algorithms (such as ChatGPT) that can be used to create new content, including audio, code, images, text, simulations, and videos.  For instance, current AI programs may use “prompts” that are short text statements to initiate the creation of images.

The Office sought to provide guidance on the registration of works containing AI-generated content, which continues to be created as new “AI Art” generators (such as DALL-E) find application in a wide variety of works. Other AI systems are maturing to even more advanced multimodal AIs that can “think” across diverse forms of data, such as vision, audio, olfactory, chemistry, code, text, images, and many others—many of which may generate works for which creators may seek copyright protection.

The Guidance attempts to clarify the distinction between copyrightable and noncopyrightable works when the work includes “expressive material,” referring to AI output that, if a human had created it, would fall within the subject matter of copyright as defined in section 102 of the Act. See the Guidance at footnote 5. 

The Guidance delineates between registrable and unregistrable works by posing the first preliminary question: “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.”

The Guidance juxtaposes two results: a 2018 denial of registrability to a visual work that the applicant described as “autonomously created by a computer algorithm running on a machine” (denied because the examiner found that the work contained no human authorship), and the recent 2023 issuance of a registration for a graphic novel  “comprised of human-authored text combined with images generated by” an AI service (even though the individual images themselves could not be protected by copyright).

The 2018 ruling seems an unremarkable denial of registrability to a work that was wholly created, rendered or “expressed” through algorithmic direction of a machine (whether that be a printer, plotter, or CNC machine). 

By contrast, the recent grant of a registration for a human-authored textual work that included AI-generated images was copyrightable, but only to the extent the text was human-authored.

The distinction between copyrightable subject matter and machine generated “works” of art and authorship is somewhat confused by the Guidance’s use of the term “expressive material” to refer to AI output that would be statutory subject matter under 17 U.S. Code § 102 but for it having been “fixed” purely through non-human machine action. AI art might be better described as having been “generated” through algorithmic sampling and assembly. In addition, use of the term “fixed” to refer to a work that is both “designed” (i.e., arrived at from a data processing standpoint), as well as physically rendered, is also somewhat confusing. 

AI-generated works may span the ranges from works that are (1) wholly of algorithmic origin (such as the Mandelbrot Set fractal arising from mathematic plots), (2) those AI-generated from prompt statements, and (3) works created by humans with the aid of AI (addressed by the Guidance), to (4) blended works co-created by humans with the “cooperative” input of AI programming.

Generative AI, of course, is well beyond the self-designed patterns of mathematic functions and has now extended to the creation/assembly of artistic works from vast data sets. See, e.g., the DALL·E 2 AI system, which generates images from text prompts and allows iterative editing of images. 

A mere AI computer prompt, like Saint-Exupéry’s Little Prince asking “S'il vout plaît... dessine-moi un mouton” (“Please, draw me a sheep”) neither makes Saint-Exupéry (nor his diminutive protagonist) the human artist that renders the drawn sheep. Of course, the development of generative AI now takes the place of the human artist by “generating” a depiction of a sheep in accordance with the data sets made available to the generative AI algorithm(s), starting only from the “idea” or brief description of a sheep in the form of a prompt. 

The more difficult question presented is to what extent “AI-enhanced” or “AI-blended” works (i.e., works of art where the original human-authored artwork is enhanced through AI-generated content in such a way that renders the two contributions somewhat inseparable) can be considered protectable by copyright. An example might be a human-sketched tree to which the generative AI program is tasked to supply leaves, flowers and coloration in accordance with prompts and algorithmic sampling and guidance. Other problematic examples might arise where a human artist and an AI computer contemporaneously and cooperatively create artwork as each follows, anticipates or otherwise “learns” from the other. This further blurs the line between human expression and machine assembly. 

If nothing else, this would make copyright applications potentially more complex in terms of the annotation required to clarify the scope of the copyright claim. See 37 CFR 201.7(c)(4). 

Human Authorship Still Required

The Guidance reaffirms that only human authorship entitles the work (or its human-authored portion) to copyright protection. 

Whether the Guidance turns upon Constitutional interpretation or public policy, the Guidance holds true to the compass that the Constitutional provision protecting copyrights was designed to protect “representatives of original intellectual conceptions of the author,” the author being human.

Recent developments of AI can certainly emulate human expression. However, all AI—regardless of the level of sophistication—involves sampling and assembly of data that only simulates human imagination and thought. However, the Guidance makes clear that when an AI technology determines the “expressive elements” of its output, the generated material is not the product of human authorship.

The Human Authorship Requirement in Application to AI-Created Works

Under the Guidance, applicants should first determine whether a work containing AI-generated material also incorporates “sufficient human authorship” to support a copyright claim. This will vary from case to case, but the Guidance indicates a consideration of whether a human has selected or arranged AI-generated material in a sufficiently creative way that “the resulting work as a whole constitutes an original work of authorship.”

The Guidance recognizes that technological tools may be part of the creative process but seems to focus on whether the creative process is assisted by—rather than wholly undertaken by—technological tools (including AI). It concludes with the touchstone that what matters is the extent to which the human had creative control over the work's expression and “actually formed” the traditional elements of authorship.

Public Policy Issues Presented by AI Authorship

Opening the door of authorship to AI-generated content would raise a wide variety of issues in the application of copyright law.

From a Constitutional standpoint, permitting machines to be immortal “authors” might to some extent skew the Constitutional intent that copyrights be granted for limited times. Copyright terms extend to a number of years beyond the life of the human author, or to a statutory term measured from the date of “creation.” Without a measurable life, only the creation-based term could apply.

Finally, the extension of copyright to wholly machine-created works could give rise to an arms race of creation as the ability of AI systems to gather, store, and generate works from a huge volume of complex data (unmanageable for humans) could lead to large copyright portfolios of artwork (as only one example) that would rapidly far surpass the ability of human artists to keep pace. 

Where from Here?

The Guidance clearly signals to applicants that only human-expressed works are registrable (and protectable), whether or not accompanied by or enhanced with computer-generated content.

To this end, each work considered for registration should be analyzed to determine how much of the creative process was that of a human versus a computer. If you instruct the computer to draw a sheep, and then go out to lunch, its likely not registrable under the current Guidance.

Accordingly, applicants would be well-advised to continue to seek the protection of copyright registration within the guard rails of the Guidance by limiting application only to human-created works, with due regard to computer-generated content or enhancements that may be noted in the application(s). This approach will best assure valid registration and the legal protection it affords. 

Of course, until court rulings clarify the copyright entitlement to computer-generated content, no doubt some applicants may still pursue a more aggressive approach to seek registration in parallel to the more constrained, human-focused approach without abandoning protection of the human-created work. 

Only the crucible of litigation will resolve the issues presented. However, it would appear unlikely that courts will depart from its earlier decisions limiting copyright to human expression—unless, of course, the judicial functions of courts are delegated to AI-driven computers that will render the decisions! 

This publication is intended for general information purposes only and does not and is not intended to constitute legal advice. The reader should consult with legal counsel to determine how laws or decisions discussed herein apply to the reader's specific circumstances.
View Full Site View Mobile Optimized