Abstract

New technologies of generative AI vastly expand the power of searches for products and services that may infringe patents. Designing such systems requires attention to the dynamics of patents and innovation, to the vast scope of potential infringers, and to the probability that any particular innovation will pose a competitive threat to a patent holder.

Existing laws allow such infringement search applications to be patented. The more interesting question is: Can an application for an infringement detector developed by AI be patented? The framework for patentability set forth in the Patent Act and specific guidance issued by the USPTO for AI inventions suggests that the answer is “yes.”

The author tested this conclusion by prompting ChatGPT to draft a patent application for a generative AI enabled infringement detector. The resulting application describes a computer program that parses a patent, identifies the relevant terms and its claims, slots them into a sophisticated and deep semantic tree, and then searches sources likely to contain indicia of products and services being offered that have features that may infringe the patent. The author reviewed ChatGPT’s draft of the complete patent application and then prompted ChatGPT to revise the application to provide for a more detailed specification, to include a Doctrine of Equivalents analysis, and to tie the invention more tightly to specialized hardware. The author then made minor revisions in language and organization to fix antecedent basis and specification support issues, created drawings, and submitted the application, along with a disclosure of how the author had interacted with ChatGPT in the drafting process, to the USPTO.

The patent examiner raised no objection or presented no rejection based on AI’s involvement in designing the invention and writing the application but ultimately issued a final rejection on 35 U.S.C. § 101 subject matter eligibility and 35 U.S.C. § 103 obviousness grounds. The experiment demonstrates how inventorship can be handled in AI-human collaboration. It also shows, however, that AI generation of patent applications increase the likelihood of obviousness rejections. The subject matter selected in this experiment invited eligibility scrutiny, so the 35 U.S.C. § 101 eligibility rejection had nothing to do with the involvement of AI in the inventive process or the prosecution of the patent application.

Keywords: Law, AI, Patents

How to Cite:

Henry H. Perritt, Jr., Patenting an AI-Generated Infringement Detector, 8 Ariz. L. J. Emerging Tech., no. 5, 2025, https://doi.org/10.2458/azlawjet.8294.