AI Summit_Sept. 13 2024

Informed by practice of its members in prosecuting, licensing and litigating patents involving the development and use of AI systems, the Section continues to view those systems as tools – either patentable themselves or in part or providing a wide range of automation but without “autonomy” to act as an inventor (or author) ( e.g., USPTO AI Comments at 3, 8, 11, and 20). Such tools include the “general” and “generative” AI systems availed of by millions of users today ( see USCO AI Comments at 3, 8, 10, 11, and 12). The Section agrees with the Office that “AI” may not be named as an inventor on a U.S. patent. However, the Section respectfully submits that the Guidance’s application of the “significant contribution” test for inventorship analysis as set forth in Pannu v. Iolab Corp , 155 F.3d 1344, 1351 (Fed. Cir. 1998), is not suitable for inventor determination on AI-assisted inventions. The Guidance improperly and unnecessarily pre-judges inventorship eligibility for human invention that should be judged, when necessary, under the fundamental “touchstone” test of (adequately complete) “conception,” which may include “recognition and appreciation” of an invention at different stages of the invention process. The Guidance casts doubt on inventorship through general acts of human direction and pre-judges lack of human inventorship in relative simple prompting of general-purpose generative AI. 1 These doubts and pre-judgment affect the innovator’s choice of availing of the patent system rather than maintaining secrecy 2 of the discovery, impacts the investment valuation, and the viability of post-issuance and litigation challenges to inventorship or adequate disclosure throughout the life of a patent. For these reasons, further explained below, the Section respectfully requests that the Office modify its Guidance in accordance with EO section 5.2(c)(ii) to consider more broadly acts of human conception, including “recognition and appreciation.” Incongruity of the Pannu Approach The Pannu factors recited in the Guidance were developed in cases in which there was no question that at least one natural person was identified as an inventor in the original patent application. Most of those cases involved the fact-finding, subject to evidentiary rules such as the requirement for corroboration and the burden of clear and convincing evidence, of particular human activity that did or did not result in the finding of that person’s inventorship, joint inventorship, or non-inventorship under the patent law. In most of these cases, inventorship of a 1 The Guidance’s rejection of inventorship through a relatively simple query with apparently straightforward results is to some degree a pre-judgment of 35 U.S.C. § 103 obviousness, which itself expressly provides: “Patentability shall not be negated by the manner in which the invention was made.” An invention’s patent eligibility also should not turn on the manner in which the invention was discovered, yet, under the Guidelines, an invention may be ineligible for patent protection if it was discovered using an AI-system, but not ineligible if discovered through trial and error (or random experimentation, chance or even by accident). The law should not treat an invention differently based on use of different tools used to discover or reduce to practice an otherwise patentable invention. 2 Although that choice is not binary, there are situations in which an inventor may delay disclosure until if and when a patent application is made or in which a patent owner may share more information than disclosed in a patent where a patent license provides compensation and conditions for such additional disclosure. The effect of limitation of actual or perceived ineligibility unnecessarily to limit patent-based knowledge diffusion would be contrary to the Office’s restatement of the view that “patents drive our nation's technological progress and achievement by ... promoting knowledge sharing to inspire others to engage in follow on innovation,” 89 Fed. Reg. 18907, 18908 (March 15, 2024) (emphasis added).

2

AI Roundtable Page 31