By Joshua Rich and Mike Borella –

On February 13, 2024, then-USPTO Director Vidal issued inventorship guidance for AI-assisted inventions;[1] on November 28, 2025, new USPTO Director Squires revoked and replaced it. But both the earlier guidance and current Revised Inventorship Guidance are based on previously decided Federal Circuit precedent. That is, the change in guidance is not based on a change in law, but rather a change of administrations. As a result, as Director Squires has emphasized on other fronts, examiners are being counseled to be more flexible and more pro-patent grant in approaching their interpretation of the same case law. That is, in sports parlance, the modification is really a “point of emphasis” – the rules haven’t changed, but officials are to focus differently on how they are to be enforced.

Both the earlier guidance and Revised Inventorship Guidance are based on 35 U.S.C. § 115 and the Federal Circuit’s decision in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), which establish that only natural persons (that is, humans) can be named inventors on patent applications or issued patents. AI systems therefore cannot be named as inventors. The question, as always, is who “conceived” of an invention, an issue that preceded the development of AI systems and continues to be a fact-intensive, challenging quandary today. In both the earlier and current guidance, examiners were directed to Burroughs Wellcome Co. v. Barr Labs., Inc., 40F.3d 1223 (Fed. Cir. 1994) for the test to answer that question.

Based on that legal background, the four major points of both guidance are the same. First, the USPTO generally presumes that the human inventors named on the application data sheet and oath and declaration are the true inventors of the invention claimed in a patent application. Second, no AI system (or other non-human person or thing) can be named as an inventor for a patent application or issued patent. Third, when multiple humans are involved in the inventive process, traditional inventorship principles (including the factors set forth in Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998)) apply. Fourth, the same analysis applies for utility patents, plant patents, and design patents.

Where the Revised Inventorship Guidance and prior guidance differ is more subtle. First, the prior guidance was published concurrently with a discussion of examples “to provide assistance to the public and examiners on the application of this guidance in specific situations.” Presumably, those examples are no longer useful in discussions with examiners.[2] Second, the Revised Inventorship Guidance rejects any reliance in the Pannu factors in determining inventorship between a human inventor and AI.[3] The prior guidance did rely on the Pannu in requiring that a human meet the test when determining inventorship on a claim-by-claim basis. And the first Pannu factor is that “a person must “contribute in some significant manner to the conception or reduction to practice of the invention,” which is always a requirement for inventorship.[4] So to the extent that there is daylight between the two sets of guidance, it is based on applying the traditional test for conception, rather than the Pannu factors, when there is only a single human inventor.

All in all, the tone of the Revised Inventorship Guidance may be more important than the changes from the previous guidance. It is far shorter and less legalistic, more in line with Director Squires’s “open for business” approach. It wasn’t published for notice and comment rulemaking, like the prior guidance, so it can more easily be revoked. Given that it relies on Federal Circuit precedent, however, that is unlikely in the absence of a change in the law. So perhaps the most important advice it provides lies in a sentence in the legal standards section: “There is no separate or modified standard for AI-assisted inventions.”

Nonetheless, a dose of caution is warranted, both for applicants and for practitioners. Just because the USPTO is unlikely to consider the inventorship during prosecution or post-grant review proceedings, this does not mean that you should not investigate carefully to ensure the proper inventors are named when the application is filed and when the patent issues. During the research process, the best practice is to keep a clean record of which features were human conceptions when AI is used in the ideation process. Challengers in court will likely scrutinize whether the human inventor(s) possessed a “definite and permanent idea” of the complete invention or merely had a “general goal or research plan” that the AI solved. If AI was used, the human inventor(s) may need to demonstrate (preferably with contemporaneously recorded evidence) that they conceived of each and every claim element themselves. Otherwise, the patent could be invalidated for incorrect inventorship or rendered unenforceable for inequitable conduct for intentional misrepresentation of inventorship.

One way for applicants to mitigate such risk is to document their specific vision of the invention, such as through sketches or written descriptions, before using an AI tool. Further, applicants should also retain records of any changes they made to designs after being generated by AI that relate to the claimed invention. Additionally, applicants should retain histories of their interactions with AI (e.g., prompt / response logs), as these records can serve as evidence that a human exercised control over the AI’s output and used the AI strictly for assistance.

Finally, the Revised Inventorship Guidance reiterates that a U.S. patent application claiming priority to a foreign patent filing must share at least one human inventor with that earlier application to be valid. When a foreign application has named an AI inventor, applicants can strip the AI from the inventorship list when entering the U.S. But the USPTO has made it clear that it will flatly reject any priority claim to a foreign application that names an AI system as the sole inventor, meaning applicants will effectively lose their earlier filing date if no human was originally named.

---

[1] For a fulsome discussion of the February 13, 2024 guidance, see https://patentdocs.org/2024/02/12/uspto-issues-guidance-about-inventorship-of-ai-assisted-inventions/.

[2] The examples were published at http://www.uspto.gov/​initiatives/​artificial-intelligence/​artificial-intelligence-resources. They are no longer available there.

[3] Indeed, the Revised Inventorship Guidance states that “Pannu is inapplicable when only one natural person is involved in developing an invention with AI assistance because AI systems are not persons and therefore cannot be ‘joint inventors’ so there is no joint inventorship question to analyze.”

[4] The other Pannu factors require a joint inventor to “(2) make 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 (3) do more than merely explain to the real inventors well-known concepts and/or the current state of the art.”