Note: This article was written and edited by humans, with outlining support from ChatGPT.
As businesses integrate AI-generated text, images, and video into their marketing strategies, a question looms: Who owns AI-generated content? For the past few years, we only had a handful of cases and some general direction from the U.S. Copyright Office (USCO) to work with. In January 2025, it issued new guidance, Copyright and Artificial Intelligence: Part 2 – Copyrightability, providing some clarity on this issue.
The report’s conclusion: AI-generated works, in most cases, are not eligible for copyright protection unless there is substantial human involvement.
Our AI consultants break down what you need to know about copyright and AI.
The new USCO guidance reaffirms a foundational principle of U.S. copyright law: only works created by humans are eligible for copyright protection. Fully AI-generated content, without significant human modification, falls into the public domain—making it free for anyone to use, copy, or distribute.
So, what constitutes meaningful human authorship? While these decisions will still be made on a case-by-case basis, the USCO has given us some direction and examples about how much human intervention is “enough.”
At this point, humans don’t have enough control over the prompt-writing process to argue that prompt engineering alone creates a copyrightable work. The report repeatedly noted several things in its analysis:
The January 2025 guidance states:
[G]iven current generally available technology, prompts alone do not provide sufficient human control to make users of an AI system the authors of the output. Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output.
This is consistent with Thaler v. Perlmutter, which involved a visual work called “A Recent Entrance to Paradise.” The image at issue was fully generated by AI without any significant human intervention other than prompt writing. The creative work was not deemed copyrightable, although this case is currently on appeal.
In its guidance, the USCO encourages us to treat AI as a tool.
In one of its examples, it references “Where That Came From,” a song that country singer Randy Travis released in 2024. Mr. Travis lives with stroke-related aphasia, which limits his speech. He and his creative team used artificial intelligence to replicate his voice, but the songwriting, production, and design of the AI model were all done and directed by humans.
Because humans exerted creative control over everything but the actual “voice,” the creative work was copyrightable.
Suppose you create an image using Midjourney or DALL-E. Then, you make significant edits to the image, adding elements and making easily identifiable creative changes. This human intervention might be enough to make your work copyrightable.
The USCO comments:
Unlike prompts alone, these tools can enable the user to control the selection and placement of individual creative elements. Whether such modifications rise to the minimum standard of originality required under Feist will depend on a case-by-case determination. In those cases where they do, the output should be copyrightable.
However, the extent of the editing will matter. For example, the USCO limited copyright protection of Zarya of the Dawn, a graphic novel. In this case, the author, Kris Kashtanova, wrote the story, created multiple Midjourney prompts for the images, made minor edits to those images, and then curated them into the graphic novel.
Initially, the Copyright Office determined that their work was fully protected. Later, the Office modified its decision, indicating that while the text and the compilation of the images were copyrightable, the images themselves were not.
Mx. Kashtanova argued that at least some of their images were edited and therefore protected. However, the Office opined that their edits were so minor as to be indetectable.
This reasoning also applies to written content. If you’re heavily reworking AI-generated copy to include your own examples, change its voice and tone, and restructure its narrative flow, the content may be protected. However, if you’re just eliminating an adjective or two, the copy will likely be considered part of the public domain.
If your marketing team is churning out AI-generated blog posts, social media copy, or visual assets, you cannot assume these works are legally protected.
AI can be used as a tool, as in the Randy Travis case, but it cannot replace human decision-making in the creative process. Cases like Thaler highlight the risks of over-relying on AI-generated content without meaningful human involvement.
RELATED: Ethical marketing and AI: Navigating challenges in highly regulated industries
Caveat: I’m not a lawyer. If you have a specific legal question or want to know how best to protect your intellectual property rights, consult with the experts. This is my interpretation of the January 2025 USCO guidance, based on my experience and research.
The Copyright Office draws a distinction between AI as a tool and AI as an independent creator. To claim copyright protection, a human must exercise “sufficient control” over the final work.
The USCO offers several indicators of human authorship that can support copyright eligibility:
To put these guidelines into context, let’s review some examples of where AI is used but copyright protection might be granted:
A human writes a novel but uses AI to refine grammar, suggest alternative phrasing, or summarize chapters. The final work is copyrightable because the core creative decisions were made by a human.
A human designer creates a mockup of an image they want to create. Next, the designer uses AI to generate a series of versions of their work. Then, they make edits to achieve their final creative vision. Because the original output was human-created and the designer exercised creative control, editing the AI-generated version, copyright protection would likely apply.
A human designer creates an image mockup, then uses AI to generate different versions of their work. The designer then makes edits to achieve their final creative vision. Because the original output was human-created and the designer exercised creative control, copyright protection would likely apply.
A human uses AI to translate their own content into other languages. (ChatGPT is quite good at translation, but always make sure you have someone proof its work.) Again, AI is being used as a tool in this case.
RELATED: Debunking the AI easy button: Why human expertise still matters
Beyond copyrightability, businesses should also be aware of other legal risks associated with AI-generated content:
To mitigate these risks, businesses should consider building in compliance by design, which proactively integrates compliance into your processes and systems from the ground up. You should also establish clear policies that direct how AI-generated materials are reviewed, modified, and used in marketing.
AI-generated content is not inherently protectable under copyright law, and your content creation should remain human directed. Here’s how you can start:
LaFleur specializes in marketing strategies that balance innovation with compliance. Whether you’re exploring AI-driven content or refining your brand’s legal protections, our team can help you navigate the evolving landscape of AI, copyright, and business strategy.
Want to discuss how AI fits into your marketing and compliance strategy? Let’s talk.
Thaler v. Perlmutter. D.C. Circuit oral argument recording (2024). Retrieved from https://www.courtlistener.com/audio/94557/stephen-thaler-v-shira-perlmutter/
Thaler v. Perlmutter. 687 F. Supp 3rd 140 (2023). Retrieved from https://caselaw.findlaw.com/court/us-dis-crt-dis-col/114916944.html
U.S. Copyright Office. (January 2025). Copyright and Artificial Intelligence, Part 2: Copyrightability. Retrieved from https://copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
U.S. Copyright Office (February 21, 2023). Zarya of the Dawn letter. Retrieved from https://www.copyright.gov/docs/zarya-of-the-dawn.pdf
Yahr, E. (2024, November 10). Randy Travis’s beautiful baritone was lost. AI helped him sing again. Washington Post. Retrieved from https://www.washingtonpost.com/entertainment/music/2024/11/10/randy-travis-artificial-intelligence-ai-sing/