By Brooke Anderson (Legal Intern)
In 2025, an age where Artificial Intelligence (“AI”) is reshaping industries – from content creation to product design to legal research – questions about ownership, authorship, and infringement of intellectual property are becoming more complex than ever. As AI becomes more creative and autonomous, one key legal dilemma emerges. Namely, can AI infringe on intellectual property (“IP”) – and if so, when, how, and who is liable?
The core three types of IP are copyright, trademark, and patents. Copyrights protect original works of authorship like songs and art. Trademarks protect brand identifiers like logos and slogans. Patents protect new, useful, and nonobvious inventions. All of these are traditionally designed to protect the creative output of humans. IP laws are provided to protect the creations of the human mind, first-in-time system for allocating entitlements, rights to exclude others from using intellectual property.
The focus of this blog will be copyright. When considering copyrights, AI can write novels, design logos, and generate code, the legal framework is under pressure to catch up to the fast-growing world of AI. Therefore, though AI systems can infringe on IP, the legal responsibility usually falls on the humans or the companies who develop, deploy, or use the AI.
AI tools like ChatGPT, Midjourney, or Sora can generate text, images, and videos that closely resemble existing works. Infringement may occur (1) when AI generates content that copies protected elements of an existing work, or (2) when AI output is substantially similar to copyrighted material it was trained on, especially if it mimics style or structure.
One of the most pressing legal questions surrounding artificial intelligence is whether its output qualifies as transformative under the fair use doctrine, or if it constitutes a derivative work that infringes on the original creator’s rights. The key legal issues emerging in this area include but are not limited to: (1) whether copyrighted material can be legally used to train AI models; (2) whether AI-generated works that include elements from preexisting content constitute fair use or infringement; and (3) whether such AI-generated works are eligible for copyright protection at all.
In Thaler v. Perlmutter, 687 F.Supp 3d 140 (D.D.C. 2023), the court upheld the U.S. Copyright Office’s decision to deny copyright protection for an artwork created entirely by an AI system without human input. The Office emphasized that copyright law protects only works of “human authorship,” a stance reinforced in a recent agency report. This issue remains active as the case proceeds on appeal and regulators continue to refine the boundaries of AI and IP law.
AI cannot own property or be sued (yet). So, liability falls on the developers who build and train the AI models, the users who prompt and deploy AI for commercial use, and the platforms that host or profit from AI’s output.
AI-generated work cannot be copyrighted at all in the United States, for now, unless there is a clear human author involved in the creation. Therefore, AI output may infringe on someone else’s copyright, but AI output may not itself be copyrightable, making enforcement and protection a tricky balance.
Businesses and creators can reduce the risk of AI IP infringement by using vetted AI tools that disclose training data sources, they can review AI-generated content before commercial use, and they can avoid generating in the style of identifiable artists or brands unless licensed.
On the developer side, the focus should be transparency in regard to the training data and the filters used. Further, developers should allow opt-outs for creators who do not want their content to be used and implement content detection tools to flag problematic outputs.
AI’s rapid evolution is outpacing existing IP law, forcing courts and lawmakers to adapt. There are no easy answers – yet – but one thing is clear: AI is not a legal loophole. Using it recklessly will not absolve a person or business from liability. It is important to stay informed, ethical, and cautious when navigating the technologies of AI and IP regulation.