Deciphering the Copyright Office’s Policy on AI Art
* Cover image By Kris Kashtanova using Midjourney AI, Public Domain, Wikimedia Commons.
The use of AI in visual art, literary works, music, and high school essays everywhere has exploded during the first few months of 2023. Uses we have already seen are only the beginning of AI capabilities for creative purposes, and yet the legal system is already running late to the AI party. Intellectual property concerns are among the obvious legal issues, and the U.S. Copyright Office (the “Office”) has released some guidance to try to address them.
In late February, the Office sent a letter to the author of a graphic novel that used AI-generated images, explaining that it would grant limited protection to the novel that extends only to the “text as well as the selection, coordination, and arrangement of the novel’s written and visual elements” but not to the AI-generated images featured in the novel. The Office reasoned that the images were not the product of human authorship, making them separate from and not included in the author’s original expression.[1]
On March 16, 2023, the Office issued a policy statement expanding on its official position on AI-generated art.[2] This statement addressed three main issues that are relevant to creators using AI-generated material. Let’s review their conclusions.
1. Is the material produced by generative AI protected by copyright?
The short answer to this question is really “no.” Content created only by AI will be ineligible for copyright protection. Current copyright law, which is based on the U.S. Constitution, the Copyright Act, and federal case law, grants protection only to works of human authorship. “Authors” are originators, meaning the work has to “owe its origin” to a person to have an author. Robots and animals are necessarily excluded from being authors.
A Deeper Dive: Artsy Animals
Do you remember Naruto, the selfie-taking crested macaque? With a photographer’s unattended camera, Naruto snapped a few photos, including his most iconic self-portrait. The animal rights organization PETA sued the photographer on behalf of Naruto, claiming the photographer violated the monkey’s copyright in the images, which the photographer had published in a book. The photographer could not claim authorship of the selfies, but as it turns out, neither could Naruto. The case was dismissed because the court found that Naruto, as a non-human animal, lacked statutory standing to sue for copyright infringement under the Copyright Act.
For more see Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018).
AI is treated the same as animals under copyright law because AI-generated works (without any creative alteration or arrangement) owe their origin to a process that operates with only minimal creative input from humans, if any.
In addition to AI-only creations, work generated from prompting by a human is also excluded by this policy since the human input does not rise to the level of creativity necessary to trigger copyright protection. So, neither you nor OpenAI can register copyrights for all of those beautiful, scary, funny, amazing images you asked DALL-E to generate.
If you use some AI-generated content within a larger body of your own work, the Office may allow registration of your contributions to the work. When an artist tries to register a work of hybrid origin, the Office must balance multiple considerations to evaluate its copyrightability. There is no bright line rule.
2. Can creators register copyrights for works consisting of both human-authored and AI-generated material?
Creators can register copyrights for works consisting of both human-authored and AI-generated material if the work is sufficiently human-authored to satisfy the Office’s standards. To determine the level of human authorship, the Office looks to answer questions such as:
Is the work basically one of human authorship, with the AI generator merely being an assisting instrument?
Are the traditional elements of authorship of the work (literary, artistic, or musical selection; elements of selection and arrangement, etc.) actually conceived and executed by a human rather than by a machine?
Are the AI contributions the result of “mechanical reproduction” or are they of the human author’s “own original mental conception, to which [the author] gave visible form”?
How does the relevant AI tool operate, and how was it used to create the final work?
If the Office finds, based on the work’s submission details, that the “traditional elements of authorship” are attributable to the AI generator, then it will not register the work. The Office will make these determinations on a case-by-case basis for each work.
It might seem like a squishy standard. Lacking precedents, this guidance might not seem concrete for artists looking to incorporate this new and exciting technology into their expression. Still, the Office claims 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.” For now, artists should look to previous decisions that have been made for other technology-assisted art and other works that incorporate non-human production. The treatment of these works will provide the best analogies for those figuring out whether their work is copyrightable.
A Deeper Dive: Prompts
Aren’t the prompts people put into AI generators pretty creative? Why wouldn’t AI-generated work be protected if the prompts are carefully engineered to produce a particular kind of art?
The Office’s current view is that creativity in prompt writing does not give users of AI ultimate creative control over the systems’ interpretation of the prompts to warrant copyright protection. They say that the prompts “influence” rather than completely determine the resulting content. The Office treats the prompts as instructions to a commissioned artist, the artist, in this case, being the AI machine that “decides” how to use those instructions to create the final result.
There are already new jobs opening up for “Prompt Engineers,” created for people who are skilled at providing precise instructions to AI to produce specific desired results. Perhaps the copyright issues will become more complicated upon considering the substantial creativity and specialized skills prompt writing involves.
3. What information should be provided to the Office by creators who want to register works containing human-authored and AI-generated material?
Artists who wish to register work containing AI-generated content must disclose the use of AI when filing with the Office. Only the human-authored contributions will be eligible for copyright protection, so the applicant must also provide a brief description of the human-authored contributions to the work. Artists will use the Standard Application, and provide this description in the “Author Created” field. Here’s an example provided by the Office:
[A]n applicant who incorporates AI-generated text into a larger textual work should claim the portions of the textual work that is human-authored. And an applicant who creatively arranges the human and non-human content within a work should fill out the ‘‘Author Created’’ field to claim: ‘‘Selection, coordination, and arrangement of [describe human-authored content] created by the author and [describe AI content] generated by artificial intelligence.’’
The use of AI-generated content in the work that is more than de minimis[3] should be excluded from the application by explicitly disclaiming it, which may be done by providing a brief description of the AI content in the “Limitation of the Claim” section in the “Other” field, under the “Material Excluded” heading.
Applications that have already been submitted without information about AI use will need to be amended, so these applicants must contact the Copyright Office’s Public Information Office to report the omitted facts. For copyrighted works that include AI-generated work that have already been registered without disclaimers, the applicant will need to correct the public record by submitting supplementary registration describing the human-authored and AI-generated contributions in the relevant fields. New certificates with disclaimers will be issued for those works containing sufficient human authorship. Consequences for failure to correct the record could include the loss of the benefits of registration.
The Office’s Next Steps
In addition to issuing this policy statement, the Office announced that it is embarking on an initiative to examine the entire body of copyright and policy issues raised by AI.[4] The Office requests input on this initiative by inviting the public to listening sessions that will take place in April and May of this year. The sessions will separately address issues related to Literary Works, Visual Works, Audiovisual Works, and Music and Sound Recordings, and interested creatives and lawyers are encouraged to attend.
Notes
[1] https://copyright.gov/docs/zarya-of-the-dawn.pdf
[2] Unless otherwise indicated, the Copyright Office’s “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence” is the source material for the remainder of this blog post. You can find this document in Vol. 88, No. 51 of the Federal Register, or at https://www.govinfo.gov/content/pkg/FR-2023-03-16/pdf/2023-05321.pdf
[3] This is a legal standard that could describe uses that are just brief quotes, short phrases, and other very limited uses of the work. If there is any uncertainty about whether the AI content is more than de minimis, consult a lawyer and/or disclaim the use just in case. The Office also notes that “[a]pplicants who are unsure of how to fill out the application may simply provide a general statement that a work contains AI-generated material. The Office will reach out to those applicants to clarify relevant issues with the use of AI.”
