No portion of this article was produced using AI. That statement carries with it a variety of interesting questions, and likely prompts many intriguing answers. Artificial Intelligence is having a very real impact on our lives, good and bad, and is pushing the boundaries of intellectual property; in particular copyrights and trademarks. Currently, there is a dichotomy between protectable elements of copyright versus trademark via AI-generated content. At a high level, copyrights are more restrictive in allowing for intellectual property protection to attach to AI-generated content. In contrast, trademarks are much more permissive for those same intellectual property protections covering AI-generated content. The rationale for this dichotomy is dictated by what each type of intellectual property is meant to protect – copyrights for works created by an author, trademarks for words, logos, or symbols that serve to identify the source of a product or service. The landscape surrounding AI and intellectual property is being constructed as we speak. While many answers remain unknown, we are seeing initial foundations being laid.

Copyrights are concerned with the creation of a unique work by an author. Based upon current legal precedent, the “author” must be human. This precedent is most easily understood through the infamous monkey selfie case, Naruto v. Slater. See Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018). The case established that a non-human lacks statutory standing under the Copyright Act to claim a copyright and sue for infringement. More recently, courts have affirmed that a work that claims AI as the sole author is also not eligible for copyright protection. See Thaler v. Perlmutter, 687 F. Supp. 3d 140, D.D.C. 2023. The foreseeable future is likely to include legal battles that seek to parse out if, and when, copyright protections attach to works that are AI-generated, or at least AI-generated in part.

The Copyright Office has already proffered some guidance in response to submissions for registration of works that are solely produced by, or include some aspect of, AI outputs. Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190. Initially published in March 2023, the Copyright Office penned guidance for prospective applicants to consider when using AI. Copyright Registration Guidance, 88 Fed. Reg. 16190. At the time of this article being written, the use of AI for solely producing an output from a prompt lacks human authorship and will not be subject to copyright registration. Id. The Copyright Office has compared this type of interaction between the human prompt and AI output as being similar to instructions provided to a commissioned artist, wherein the instructor provides what they wish to have completed, and the commissioned artist determines the actual expressive output of the final work. Id. at 16192. But that does not completely shut the door on having a copyright claim that includes AI-generated content. If the work contains “sufficient human authorship” then it may still be eligible for copyright, but only to those portions of the work that are human-authored. Id. As an example, selection or arrangement of AI-generated content may constitute sufficient human authorship to protect the selection or arrangement, but not the content itself. Id. Likewise, modification of AI-generated content may also meet copyright protection if the modification is sufficient to rise to the level of human authorship. Id. As provided in the Copyright Office’s guidance, “a visual artist who uses Adobe Photoshop to edit an image remains the author of the modified image.” Id. at 16193. But an artist who uses the AI-powered generation feature of Adobe Photoshop will be subject to the analysis of “sufficient human authorship” in determining copyright eligibility.

Due to these issues of copyright protection, or lack thereof, there is a sense of uncertainty on how best to utilize AI-generated materials, if at all. To the extent that a “work” is created, such as a video, an image used in marketing material, text generated for a website, etc., there is a legitimate concern that such “works” when produced with AI are not protectable. When entering into a work-for-hire agreement, it is becoming increasingly common to have a section on the use, or non-use, of AI-generated content in the work. For some, there is a flat-out prohibition on any use of AI in the creation of the work, thus preventing any issues from arising as to the protectability of the work that is created. In other situations, where protectability is less of a concern, AI-sourced material provides a distinct advantage in time and cost for generating content.

Turning now to trademarks, there is a noticeable difference in the use and acceptance of AI-generated images for branding, but diligence must still be applied to the use of such content. Trademarks gain protection when a word, phrase, or logo is attached to a good or service that enters the stream of commerce. The “creator” of such word, phrase, or logo is of no consequence to gain trademark protections. Thus, the use of AI-generated content for trademarks does not induce the same level of risk when compared to ownership of copyright protection. The U.S. Patent and Trademark Office issued guidance in April of 2024 regarding the use of AI in preparing and prosecuting trademark applications. Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office, 89 Fed. Reg. 25609. No specific comments were made regarding the rejection of trademark applications that use AI-generated material as the mark. Instead, the USPTO reiterated a practitioner’s obligations to use oversight and diligence in producing AI content, especially when using AI tools for preparing documents or other filings with the Office. Thus, at this time, utilizing AI to generate a potential trademark for filing at the USPTO will not draw any specific rejection based on the AI-generated content alone.

While trademark protection is one issue for consideration, a second issue must also be taken into account. That issue is whether the AI-generated content is infringing on another’s trademark. For example, prompting an AI tool to create ten brand names for tires specifically designed for electric vehicles (EVs), may result in outputs that are already trademarked by other companies. This is why due diligence must still be performed to oversee the output of the AI tool and assess any potential infringement liability. There are no guarantees that the AI-generated content is free of trademark infringement; or copyright infringement for that matter. Before implementing a branding strategy, a review of trademarks to assess any potential likelihood of infringement should be undertaken, just as if the branding content was “human authored.” Such risks are prompting companies who license others’ IP to include indemnification clauses requiring the licensor to indemnify the licensee against claims of infringement for content that was obtained from a generative AI system.

There are currently more than a dozen legal cases pending across the U.S. that are grappling with the intersection of AI technology and intellectual property. Many dealing with the use of copyrighted material in training AI models and how that affects the copyright holder’s rights. A growing number of supporters are encouraging use of “ethical” AI models that are trained on open-source material and exclude protected material or compensate the IP-holder for use of the material. By using “ethical” AI models, potential liability may be reduced for AI-generated content that uses infringing source material, but that fight is ongoing in the courts and yet to be resolved. The outcome of these cases, with their own relevant factual findings, will likely begin to parse out some of the issues related to copyrights and trademarks, allowing us to make more informed decisions on how and when use of AI tools is appropriate.