On November 28, the United States Patent and Trademark Office (USPTO) issued a notice titled Revised Inventorship Guidance for AI-Assisted Inventions (“the Revised Guidance”), expressly rescinding its original guidance on that subject published in February of 2024 (“the Original Guidance”). As previously discussed, the Original Guidance followed case law holding that artificial intelligence (AI) cannot be a named inventor in a patent application, but allowed use of AI as a tool to develop an invention, so long as a human inventor made a “significant contribution” to that development.

In the case law realm, the overruling of earlier precedent usually produces a significant change in the law. One might expect a similar result in this “rescinded in its entirety” USPTO policy context. The USPTO jettisoned the catchphrase “significant contribution” in favor of traditional inventorship principles that it concludes universally apply, regardless of whether AI played any role in the invention development. Close scrutiny, however, reveals that this amounts to a distinction without a practical difference, either to inventors or to the patent practitioners representing them.

Inventors using AI to assist with invention development should therefore continue to maintain records of their prompts and the corresponding AI outputs to prove human inventorship, should an inventorship inquiry arise either during prosecution of the patent application in the USPTO or during litigation over any issued patent. Moreover, none of the USPTO’s inventorship guidance is binding on federal courts, which have yet to consider the issue of inventorship in the context of AI-assisted invention development. This judicial uncertainty counsels in favor of patent practitioners continuing to inquire about the use of AI by their inventor clients, to minimize chances that any inequitable conduct issue could ever arise in litigation over an inventorship question.

The Original Guidance announced that “each claim must have been invented by at least one named inventor. In other words, a natural person must have significantly contributed to each claim in a patent application or patent.” The USPTO derived the “significant contribution” standard from a Federal Circuit decision titled Pannu v. Iolab Corp.,155 F.3d 1344 (Fed. Cir. 1998).

The Original Guidance acknowledged the USPTO policy that usually, the USPTO presumes that the inventors named in a patent application are the correct inventors. However, the Original Guidance cautioned that if a patent examiner determines from the application file or from other evidence that any claim was not invented by a natural person, to reject any such claim under 35 U.S.C. § 101 and 115, the latter section requiring that a patent application include the name of the inventor or each joint inventor.

To help define the boundaries of what could be considered a “significant contribution” by a human inventor, the Original Guidance listed four principles. For example, under one of those principles, merely taking the output of an AI platform and reducing it to practice would not likely meet the “significant contribution” standard. “However,” noted the Original Guidance, “a significant contribution could be shown by the way the person constructs the prompt in view of a specific problem to elicit a particular solution from the AI system.”

Of particular interest to patent practitioners, i.e., patent attorneys and agents, the Original Guidance also imposed new obligations to inquire about the client inventor’s use of AI, if any, in developing the invention. The Original Guidance warned: “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 . . . .”

The Revised Guidance rescinded the Original Guidance because, according to the USPTO, Pannu only applies to human joint inventorship assessments, and because under Thaler v. Vidal, 43 F.4th 120 (Fed. Cir. 2022), AI cannot be a joint inventor. The USPTO therefore suggests (without expressly stating) that the Original Guidance erred by using Pannu to derive the “significant contribution” standard for situations involving AI-assisted inventions.

“There is no separate or modified standard for AI-assisted inventions,” declares the Revised Guidance. In place of the now-discarded  “significant contribution” standard, inventorship determinations for AI-assisted inventions will now be done according to traditional inventorship notions of “conception.” Articulating the traditional standard, the Revised Guidelines explain: “The question is whether the natural person possessed knowledge of all the limitations of the claimed invention such that it is so ‘clearly defined in the inventor’s mind that only ordinary skill would be necessary to reduce the invention to practice, without extensive research or experimentation.’” Thus, the Revised Guidelines instruct: “When one natural person is involved in creating an invention with the assistance of AI, the inquiry is whether that person conceived the invention under the traditional conception standard . . . .”

As a practical matter, is this traditional “conception” inquiry in the AI-assisted context all that different from the discarded “significant contribution” standard? Not really.

In the AI-assisted context, ascertaining whether a natural person acquired a level of knowledge of the invention such that “only ordinary skill would be necessary to reduce the invention to practice, without extensive research or experimentation” basically comes down to whether the natural person, as opposed to the AI tool, was the source of the novel concepts of the invention. If the natural person relied upon AI to derive one or more of those concepts, then considerable doubt can arise as to whether the natural person satisfied the traditional “conception” standard.

Similarly, if a natural person did not construct an AI prompt “to elicit a particular solution from the AI system” (per the Original Guidance), this suggests that the natural person did not make a “significant contribution” to the development of the invention. It likewise indicates that this natural person did not possess a level of knowledge of the invention sufficient to construct informed AI prompts. The lack of informed AI prompting can thus lead to a failure of this natural person to attain inventor status under both the traditional “conception” standard and the discarded “significant contribution” standard.

Consequently, despite the USPTO’s recission of the Original Guidance, the four principles listed in the Original Guidance remain a useful reference for guiding the inquiry of whether a natural person in an AI-assisted scenario meets the traditional “conception” standard of inventorship. This is especially so since the Revised Guidance lacks any recitation of principles providing the contours of when a human being who used AI to help develop an invention can meet the traditional “conception” standard. As a result, it is still a good idea to maintain and preserve records of AI prompts and outputs generated during the development of an AI-assisted invention.

Nor does the recision of the Original Guidance mean that patent practitioners can now simply avoid questioning a putative (would-be) inventor about the use of AI in the invention development process. Since, for the reasons discussed above, the construction of AI prompts informs the inquiry of whether a natural person meets the traditional “conception” standard, questions concerning AI use and prompting remain relevant to determining whether the putative inventor is in fact an “inventor” under the Patent Act.

Furthermore, the Original Guidance recited a critically important disclaimer: “This guidance does not constitute substantive rulemaking and does not have the force and effect of law. The guidance sets out agency policy with respect to the USPTO’s interpretation of the inventorship requirements of the Patent Act . . . . ” Although the Revised Guidance does not repeat this disclaimer, it carries the same limitations, since it is just an updated USPTO policy statement.

Even if the Revised Guidance were deemed to rise to the level of rulemaking, federal courts could still choose to disregard it. In the landmark case of Therasense, Inc. v. Becton, Dickinson and Co., 649 F.3d 1276 (Fed. Cir. 2011), the Federal Circuit decided the issue of which standard of materiality governs the question of whether “material information” was withheld from the USPTO for purposes of establishing an inequitable conduct defense to a charge of patent infringement. Different iterations of a USPTO regulation (37 C.F.R. § 1.56) recited different rule-based standards of materiality, while Supreme Court precedent suggested a “but-for” materiality standard, i.e., “prior art is but-for material if the PTO would not have allowed a claim had it been aware of the un-disclosed prior art.” In adopting the “but-for” standard, the Federal Circuit concluded that, in view of the USPTO’s evolving materiality standards: “Experience thus counsels against this court abdicating its responsibility to determine the boundaries for inequitable conduct.”

A federal court considering whether to follow the Revised Guidance as an inventorship standard for AI-assisted inventions could very well cite Therasense, as well as the disclaimer in the Original Guidance, to justify disregarding the Revised Guidance. With courts not having yet decided whether traditional “conception” standards govern the inventorship question for AI-assisted inventions, considerable judicial uncertainty presently pervades this issue.

Given the uncertainty surrounding the question of what standard would govern in a litigated challenge to inventorship involving an asserted patent claiming an AI-assisted invention, named inventors should continue to preserve their invention development records, especially including AI prompts and corresponding AI outputs. By the same token, patent practitioners should continue to inquire about the use of any AI in the development of their clients’ inventions.