by Dennis Crouch
The U.S. Patent and Trademark Office (USPTO) recently published examination guidance and a request for comments on the treatment of inventorship for inventions created with the assistance of artificial intelligence (AI) systems. Inventorship Guidance for AI-Assisted Inventions.
The key takeaway here is that the USPTO believes that an AI-developed invention is patentable so long as a human satisfies the joint-inventorship standard of “significantly contributing to the invention.” A human who provides a significant contribution may be the sole inventor and original owner, even in situations where the AI provided the greater contribution.
The PTO’s approach here is fairly broad and will likely serve current AI use cases in most situations because most AI invention models of today are tightly controlled and managed by humans rather than simply arising from AI autogeneration or broad prompting. Thus, from a practical effect, there will likely be at least one natural person who satisfies the joint inventorship standard in the vast majority of cases. (Recognizing here that mere control of an AI is insufficient. Rather, the human must provide significant inventive contribution).
Although I am very sympathetic to recognizing human contributions, I also want a patent system that broadly encourages innovation without either prescribing or proscribing particular approaches. The flexibility of this guidance allows room for both human and machine intelligence to intersect in the creative process. But, we should continue monitoring the effects the policy to ensure it does not unduly constrain AI’s eventual capacity to autonomously formulate inventive concepts. But for now, the USPTO’s basic framework reasonably balances competing interests.
But, the USPTO’s approach is not fully grounded in the law because it allows for patenting of an invention in a situation where no human or combination of humans fully conceived of and originated the invention. Rather, we are simply looking for at least one human who provided a significant contribution. The guidance does not particularly address this issue and, by declining to specifically justify the legal grounds why human “significant contributions” suffice even without complete conception, the USPTO leaves the door open to contrary arguments. Opponents could contend that full conception remains legally required for inventorship and that this expansion of the inventorship doctrine exceeds the statutory language. It is not clear who will have standing to make this particular argument.
No AI Inventors Allowed
First, the notice recognizes and follows existing case law that only natural persons can be listed as inventors on U.S. patents and patent applications. This principle was established in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), where the Federal Circuit held that the term “individuals” in the Patent Act refers only to human beings barring. Consequently, an AI system cannot be named as the inventor on a patent application, even if it was instrumental in developing the claimed invention.
The Law Requires Significant Contribution at Least One Human
The USPTO guidance explains that just because an AI system assists with an invention does not automatically render it unpatentable. Rather, for an AI-assisted invention created by a natural person or team of people working with an AI system, the key analysis focuses on whether those individuals made “significant contributions” to the conception of the invention. In other words, in order for an invention to be patentable, there must be at least one human who made “significant contributions” to the creation of the invention — either conception or reduction to practice.
Drawing from existing case law on joint inventorship, the notice refers to three main factors, known as the “Pannu factors,” for evaluating significant contribution:
- contributing “in some significant manner to the conception or reduction to practice of the invention”,
- making “a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention”, and
- doing “more than merely explain to the real inventors well-known concepts and/or the current state of the art”
Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998).
The Pannu factors are derived from the situation of joint-inventorship among humans, where multiple individuals each provide some contribution. Here, the approach is slightly different — effectively joint inventorship among human and AI contributors. In the analysis, the AI can never be considered an inventor regardless of its level of contribution. The human will be an inventor so long as they would be considered at least a joint inventor.
Typically for joint inventorship, we first look on a claim-by-claim basis to determine the inventors, but then we list inventors at the whole patent level. Thus, in a patent with three joint inventors it could be that one of the inventors only contributed to the subject matter found in one of the claims. The human-AI collaboration is different. According to the USPTO, each claim needs to have a human inventor. Claims without a human inventor should be cancelled. This ensures that every surviving claim has at least one natural person satisfying the Pannu factors.
One query is involves dependent claims. For instance, imagine an application with Claim 1 covering a method of predicting protein folding stability using a jointly-developed human-AI technique, along with dependent Claim 2 adding an additional limitation associated with predicting a specific family of proteins conceived purely by the AI system. Here, the human inventor likely deserves credit for the core concept in independent Claim 1, so long as they provided a significant contribution to its conception, including utilization of the AI system. Although the AI alone conceived the further embellishment in Claim 2, a reasonable argument can be made that the human inventor still made an overall contribution to Claim 2 by virtue of its dependency on Claim 1. Accordingly, the human could potentially qualify as an inventor on both claims. Nonetheless, reasonable minds could differ on this conclusion for dependent claims, and applicants should carefully assess inventorship on both independent and dependent claims under the new USPTO guidance.
Principles for AI Inventorship Analysis
Applying these factors to AI-assisted inventions can be challenging. Accordingly, the notice sets forth several guiding principles, including:
- AI Usage Valid: Using an AI system does not negate a natural person’s possible inventor status, provided they meet the Pannu factors. Example: A researcher who uses a protein-folding AI tool to help design a new enzyme can be listed as the inventor because they formulated prompts for the AI and then used their expertise to select the most promising enzyme outputs for further testing.
- Prompting with the Problem Insufficient: Merely presenting a problem to an AI system or recognizing an invention from its output is likely insufficient for inventorship. Example: An engineer asks broadly asks an AI system to provide an improved product design that solves a known problem in the existing product. Merely presenting the problem to a general AI without further conception contributions from the engineer would likely fall short of establishing inventorship.
- Mere RTP Insufficient: Reducing an invention to practice is typically not a significant contribution to the invention, unless the process is beyond the level of ordinary skill. Example: An engineer reduces to practice an AI system’s independently conceived invention by building and testing a prototype device. However, the programmer had no further role in developing the inventive concept beyond recognizing the AI’s output and constructing the prototype using routine skills. This reduction to practice alone likely does not rise to a significant contribution for inventorship.
- Essential Development Sufficient: Developing essential components like datasets or model architectures that are designed to address a specific problem and that help the AI system produce the invention may constitute a significant contribution. Example: A computer scientist trains an AI system on a specifically curated dataset and novel neural architecture so it can output a more efficient fuel injector design. These essential developmental contributions leading directly to the AI’s inventive output could allow the scientist to qualify as an inventor on the generated invention, although this example might be questionable.
- Mere Control Insufficient: Owning or maintaining control over the AI system alone does not make someone an inventor. Example: An engineer purchases an off-the-shelf generative design AI system and simply runs it on a computer cluster under her oversight. Despite managing the AI system, if the engineer provides no further conception input or enhancement of the system or its output, mere ownership and control would be deemed an insufficient contribution for inventorship.
Although these are some examples that serve as guiding principles. The Guidance (and the law) repeatedly state that inventorship is a highly fact-intensive question that has to be analyzed on a case-by-case basis. One note – Although I derived the principles from the guidance, my particular examples are mine (Crouch) and did not come from the guidance itself.
Additional Impacts on Patent Practice
Besides providing a framework for inventorship evaluation, the USPTO notice covers several other issues like duties of disclosure and priority claims for AI-related applications. Of note, examiners can request information about inventorship even if not necessarily material to patentability. Typically, an examiner will not formally request information absent “a reasonable basis to conclude that one or more named inventors may not have contributed significantly to the claimed subject matter.” In addition, the USPTO believes that patent attorneys and agents have a continuing duty of reasonable inquiry into the inventorship that begins with submission of the patent application. “Failing to inquire when the circumstances warrant such an inquiry may jeopardize the validity of the application or document, or the validity or enforceability of any patent or certificate resulting therefrom, and could result in sanctions or other actions under 37 CFR 11.18(c).”
Seeking Public Comments: The USPTO is seeking public comments the inventorship guidance, and has set a 90 day deadline. Meanwhile, the guidance appears to immediately impact inventorship evaluations for patent applications.