From text generators like Chat GPT to AI art generators like DALL·E, the recent explosion in the popularity of AI tools has led to creators exploring ways to incorporate AI tools into their work. But this opens them up to a whole host of legal issues.
As previously reported by Passionfruit, creators have become divided over the use of AI in making art, with several online artists claiming that the algorithm stole and reproduced artists without credit in order to produce its creations.
And after a few months of speculation, the law is sparking new debate among creators, with the U.S. Copyright Office releasing a letter explaining their decision to cancel a portion of the copyright registration it previously issued for Kristina Kashtanova’s graphic novel, Zarya of the Dawn.
The graphic novel first hit headlines in September 2022 after becoming the first AI-generated work to receive a copyright certificate. Kashtanova created the graphic novel’s artwork using Midjourney’s text-to-image AI tool, as they publicly documented on Instagram and Facebook in 2022.
The Copyright Office, which was not aware of this at the time of granting the graphic novel’s certificate, soon began an investigation into the creative process upon discovering this information.
Now, the Copyright Office confirms in its letter the use of an AI tool fails to meet the human authorship requirement under U.S. copyright law. The result is that any AI-generated materials would be considered public domain.
“The process by which a Midjourney user obtains an ultimate satisfactory image through the tool,” the Copyright Office explains in the letter, “is not the same as that of a human artist, writer, or photographer.”
Additionally, the Copyright Office agrees the text of the story and the selection and arrangement of the text and images into a graphic novel format, referred to as compilations, are protected by copyright.
“This is a great day for everyone that is creating using Midjourney and other tools,” Kashtanova wrote in a tweet and statement to Passionfruit regarding the ability to maintain protection over their arrangement and story elements. “My lawyers are looking at our options to further explain to the Copyright Office how individual images produced by Midjourney are a direct expression of my creativity and therefore copyrightable,” they confirmed separately.
The public record for the copyright certificate will be updated, and a new certificate will be issued, to reflect the registration covers only those copyrightable elements.
While registration of a work with the Copyright Office offers additional benefits and protections for a creator, registration is not a requirement to establish copyright.
“Registrations should clearly delineate the aspects of the work that were the subject of creative (human) expression,” copyright expert and attorney Aaron Moss told Passionfruit. “Unless you’re actively looking to end up as the next test case, you should make the copyright examiners’ jobs easier. Believe it or not, they’re trying to help you!”
However, the fact that generative art has now been explicitly identified as not copyrightable for lack of human creation, it doesn’t matter, legally, whether or not a creator seeks to register their AI-generated content or not.
How does this impact Zarya of the Dawn?
A potential issue for Kashtanova is their use of an AI tool to create the visual look of the titular character. This is because copyright law allows for the protection of the drawings, pictures, and descriptions of a character.
However, that protection is potentially at risk for character images and similar depictions that are generated by an AI tool, absent future insights or guidance from the Copyright Office.
“If I was Mx. Kashtanova, this would be my primary concern,” noted IP attorneys Ash Kernen and Eliana Torres in a joint Twitter thread. “A given image can be replaced, but a character guides the entire ecosystem.”
For Zarya of the Dawn, Kashtanova was also very public about using references to actress Zendaya as part of the prompt engineering process. Such actions potentially open the door for additional issues beyond copyright, such as violations of the right of privacy and the rights of name, image, and likeness (NIL).
“The future of creativity depends on creators being able to distinguish what is an appropriate use of AI from what is not,” Hannah Peterson, founder of AI Daily, told Passionfruit. “Usually we have regulations to guide us, but right now we’re in a legal gray area. We have to think really critically about what’s ethically sound and responsible until the laws catch up.”
What can creators learn from this?
“It’s important to learn about [AI] in a balanced way,” Peterson told Passionfruit. “Make sure you’re not only focusing on all the ways it can make your life easier but also really being honest with yourself about the risks you’re taking on.”
If a creator is using AI tools as part of their production process, the most important question becomes the amount of human authorship, if any, that can be attributed to the final work. But this might be difficult until the Copyright Office releases more explicit guidance about how much human authorship is required for an AI-generated work to qualify for copyright protection.
In the meantime, Adobe Stock, a stock asset licensing service through which Kashtanova licenses generative art, recently released its own set of requirements for the submission of generative art that creators may find helpful.
A potential starting point for creators that are using AI-powered tools in their workflow is to ask: “What is the end goal for this content?” If a creator is building evergreen content, such as a course, blog, or book, then it may be worth limiting the incorporation of AI content. If a creator is trying to maximize their workflow to produce a high volume of content, such as a video for TikTok, without concern regarding the ability to protect the content over the long term, then it may not matter as much.
Creators can also identify the most important elements of their work that they would want to protect, and then, ensure those elements include as much organic creative direction and as little AI involvement as possible.
Both creators and their brand partners should be working closely together if AI or similar generative tools are part of a production workflow for advertisements. The key is to ensure transparency and disclose to any partners from the beginning if you’re using AI in your product and how.
How will this case impact creator-oriented technology?
Many creator-focused software companies are expanding their offerings to incorporate AI capabilities, often powered by third-party startups such as OpenAI. For example, Kajabi’s AI Creator Tools product helps creators generate course outlines, email content, landing pages, and course lessons.
“The hardest part about content creation is simply getting started and AI is solving this problem in a big way for creators,” Vinay Mysoor, VP of Corporate Development and Legal at Kajabi, told Passionfruit. “[Creators are] using the AI outputs for inspiration and then modifying them to make the content their own.”
However, tools like this could potentially spark legal debate. A creator that uses the course outline tool, for example, may be able to claim ownership of a final course and the content within any such course, since there would arguably be significant human creative involvement to get from an outline to a more robust course offering.
However, the other three tools that can be used to output entire works may cause issues if used without any additional human involvement. Kajabi explained in a written statement to Passionfruit it does not claim ownership rights over the final course output and leaves that for creators, pointing to its Terms of Service. The company also points to OpenAI’s terms that align with that stance.
For Kajabi, the future involves making AI an accessible tool for creators. “Rather than backing away from innovation out of fear,” Mysoor said, “we’re empowering creators to unlock the good that can come from AI, which in our case is helping them turn their knowledge into thriving online businesses.”
Another example of an AI tool for creators is Revoice, a text-to-speech algorithm released by podcasting platform Podcastle that is tailored to a specific host, actor, or personality’s voiceprint—what the company calls a “Digital Voice.”
According to the website, creators read 70 sentences to train the AI model and are then able to generate anything, “from podcast intros to ad segments, voiceovers to entire episodes.”
“Creators who use Podcastle are unequivocally the owners of their generated content, using Revoice or any other tool,” a spokesperson for Podcastle told Passionfruit in a statement. “Podcastle is a content creation platform, and we do not own any of the content produced by our users.”
However, the legal terms and privacy policies that are agreed to when creators use a production platform only cover the relationship between those two parties and don’t override protections that may or may not be available under current copyright laws.
Text-to-content tools arguably now fall within a legal gray area based on the Copyright Office’s analysis in its letter. When using tools that output content of any kind, like Revoice, the content created as the input, such as a written script, may be copyrightable, but it is currently unclear what, if any, rights the creator will have to the output.
Does the resulting audio file generated by an algorithm qualify as sufficient human involvement? Does it even matter if the underlying content, such as the script, can be protected and used as part of a creator’s IP strategy?
The Copyright Office’s letter may not be the end of the road for obtaining protections over AI-generated works, as it may likely face a challenge in court. The Copyright Office arguably leaves the door open to potentially obtaining some level of protection. However, creators and brands should proceed with caution while using AI-powered tools until specific guidance is released or laws change to recognize non-human authors.