AI: Tool or Inventor? – Inventorship Pitfalls When Using AI in R&D

PDF

Generative AI is a subset of artificial intelligence that enables machines to generate content, such as words, images, voice, and even synthetic data. While Generative AI is relatively new technology, its origins from text-based analysis, to rule based analysis, and natural language processing has been around for decades. The present-day Large Language Models (“LLMs”) have the potential to augment human behavior beyond what was previously available with human to computer interaction. This augmentation is currently being used to expand human creativity, extract insights, provide answers, and identify problems. LLMs are able to do so by synthesizing trillions of data points and following patterns in the existing data.

This leads to a peculiar question, can an LLM be an inventor, especially in light that it is only following patterns in existing data. Presently, there is an emerging legal question around the use of generative AI: at what point is the AI no longer a tool but the inventor?

The Federal Circuit, in Thaler v. Vidal, ruled that an AI agent cannot be listed as an inventor on U.S. Patent Applications; a natural human must be the inventor.[1] This then raises the question of how much can an AI do before it becomes the inventor? While courts are still working to develop a clear answer to this question, the U.S. Patent Office (“PTO”) has released several notices in the Federal Register that provide guidance on inventorship for AI-assisted invention.[2] This guidance by the PTO helps shed light on some practical steps that companies and inventors can take to help ensure that a human is inventing their inventions.

At the outset, the PTO has made clear that, “the use of an AI system by a natural person(s) does not preclude a natural person(s) from qualifying as an inventor (or joint inventors) if the natural person(s) significantly contributed to the claimed invention.”[3] We are then left grappling with the age old question of what comprises significant contribution, and a study on the long accepted Pannu factors. [4] The Pannu factors provide for a weighting of conception, quality of contribution, and explanation of the state of the art to determine whether a person qualifies as an inventor.[5]

The first Pannu factor requires that a human conceive of the invention. Thus, the generative AI usage should be limited to simply following instructions provided by the human, versus creating the core concept. Prompting may matter, and we caution against deriving a solution from prompting. Rather, having conceived of the invention, a natural person may supplement their understanding and reduce the invention to practice through generative AI. It is therefore important, in case inventorship later becomes an issue, for applicants to keep a precise record of the prompts supplied to a generative AI agent – especially if the conception of the invention was derived by a human and refinement was conducted through generative AI prompting.

The second Pannu factor is whether the alleged inventor made a contribution to the claimed invention that is not insignificant in quality, when measured against the dimension of the full invention.[6] This second factor is subjective in the sense that the inventor must have made a meaningful impact, something more than routine experimentation.

The third factor is whether the alleged inventor did “more than merely explain to the real inventors well-known concepts and/or the current state of the art.”[7] This factor is similar to the conception factor in that the deciding point is whether a human conceived the initial idea that evolved into the invention. If the human was the one who reduced the invention to practice after it was conceived by an AI, this will indicate that the human is not an inventor.

The PTO has issued additional guidance in light of the Pannu factors and similar case law, of which the PTO notes that “[m]erely recognizing a problem or having a general goal or research plan to pursue does not rise to the level of conception.”[8] The PTO has elaborated that merely presenting a problem to an AI will not qualify a natural person as the inventor.[9] Further a human recognizing the output of an AI as an invention will not qualify that individual as an inventor.[10]

Lastly, and most importantly, “[a] natural person who develops an essential building block from which the claimed invention is derived may be considered to have provided a significant contribution.”[11] An example to this is a situation in which a person designs, builds, or trains an LLM in view of a specific problem to elicit or verify a particular solution.[12]

We end this article with a stark reminder that the PTO’s rules and guidance can be overridden by: the Office of the President, Congress, and the courts.

^{[1] Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022); see also Christina Nieves, Stephen Thaler v. Katherine K. Vidal 43 F.4th 1207 (Fed. Cir. 2022), 27 U.S.F. Intell. Prop. & Tech. L.J. 85 (2022).
[2] E.g., Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. 10043 (Feb. 13, 2024).
[3] Id. at 10046.
[4] Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. at 10047; Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998).
[5] Id.
[6] Id.
[7] Id.
[8] Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. at 10047 (citing Burroughs Wellcome Co. v. Barr Lab'ys, Inc., 40 F.3d 1223, 1224 (Fed. Cir. 1994)).
[9] Id.}
[10] Id.
^{[11] Id. at 10048
[12] Id.}