The Unted States Patent and Trademark Office (USPTO) recently published new guidance on subject-matter eligibility as related to Artificial Intelligence (AI), opening a written comment window to respond with a deadline of September 16, 2024. The guidance was brought about by an October 2023 Executive Order on the “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence,” in which the White House stated that “[p]romoting responsible innovation, competition, and collaboration will allow the United States to lead in AI and unlock the technology’s potential to solve some of society’s most difficult challenges,” which “requires investments in AI-related education, training, development, research, and capacity, while simultaneously tackling novel intellectual property (IP) questions and other problems to protect inventors and creators.” Executive Order at Sec. 2(b).

In this newly released guidance, the Patent Office seeks to “assist USPTO personnel and stakeholders in evaluating the subject matter eligibility of claims in patent applications and patents involving inventions related to AI technology (AI inventions),” as well as “announce[] a new set of examples that are intended to assist USPTO personnel in applying the USPTO’s subject matter eligibility guidance to AI inventions during patent examination, appeal, and post-grant proceedings.” Guidance at Summary. As before, the USPTO remains committed to issuing patents comprising AI-related inventions, provided that the inventions follow the established guidelines for patent eligibility (namely under 35 U.S.C. 101), which the Office now looks to clarify in view of the rapidly evolving AI landscape.

Background

The USPTO has recently and consistently evaluated whether its existing procedures are sufficient to handle a rise in AI-related applications; in October 2020, despite concern amongst some commenters that AI inventions are at risk under the existing subject matter eligibility analysis due to a risk of improper characterization as abstract ideas, the USPTO published a report titled “Public Views on Artificial Intelligence and Intellectual Property Policy.” The report took a comprehensive look at the stakeholder feedback received in response to the questions posed in its previous August 2019 notice, and concluded that “[a] majority of commenters agreed that AI is viewed best as a subset of computer-implemented inventions. Therefore, this majority felt that current USPTO guidance, especially on patent subject matter eligibility and disclosure of computer-implemented inventions, is equipped to handle advances in AI.” Guidance at A (emphasis added).

Since then, the USPTO has continued to seek comment on, and provide Federal Notices regarding, AI-assisted inventions, including in its 2024 Inventorship Guidance issuance, which confirmed the Court’s holding “that only a natural person can be an inventor, so AI cannot be.” Thaler v. Vidal, 43 F.4th 1207, 1213 (Fed. Cir. 2022), cert denied, 143 S. Ct. 1783 (2023), and reiterated 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.” Inventorship Guidance for AI-Assisted Inventions, February 2024, U.S. Patent and Trademark Office (emphasis added). Accordingly, it has been USPTO practice that while AI cannot be an inventor, AI systems are not precluded by assisting in the invention of a technology, provided “significant[] contribution” from the natural person inventors.

Moreover, the 2019 Revised Patent Subject Matter Eligibility Guidance” (2019 PEG) and October 2019 Patent Eligibility Guidance Update” (October 2019 Update) released by the USPTO, subsequently incorporated into the Manual of Patent Examining Procedure (MPEP) at sections 2103-2106.07, provided revised procedures for Examiners to identify abstract ideas and determine patent eligibility. The Office reported that, following the issuance of these documents, a study undertaken by the Office of the Chief Economist found that the 2019 revisions to the eligibility guidance resulted in a 25% decrease in the likelihood of Alice-affected technologies (including AI) receiving a first office action with a rejection for patent ineligible subject matter, and a 44% decrease in “uncertainty regarding determinations of subject matter eligibility” for the relevant technologies over the previous year. Guidance at B. Thus, and despite ongoing uncertainty regarding inventorship and patentability, there has been no shortage of USPTO guidance regarding AI-related inventions.

Alice/Mayo Guidance

Under 35 U.S.C. 101, which defines patent-eligible subject matter, an invention must fall into one of four categories: processes, machines, manufactures, and compositions of matter. Aiding in such a determination is the two-step test established in Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014), and Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66 (2012). This USPTO guidance thus seeks to provide further color, in view of the evolving AI landscape, to both subparts or “prongs” of Step 2A of the subject matter eligibility analysis under Alice and Mayo, as described below.

Step 1 of the analysis asks whether the claim is directed to one of the four categories of patentable subject matter. If so, and the eligibility of the claim is self-evident (for example, a pharmaceutical composition claim), then the claim qualifies as patentable subject matter without further analysis. If Step 1 is answered affirmatively, but the patentability is not self-evident, the second step (Step 2), taken from Alice, examines whether the claim recites such an inventive concept as to make the claim eligible for patenting.

Specifically, under Step 2, an Examiner must first evaluate whether a claim recites an abstract idea or other judicial exception (Step 2A, Prong One). Guidance at Background, II; see also MPEP 2106. If the claim does not recite a judicial exception, it is considered eligible, and the eligibility analysis ends. If the claim recites an exception, the Examiner must evaluate whether the claim recites additional elements that integrate the judicial exception into a practical application (Step 2A, Prong Two). Id. Courts have consistently found abstract ideas, laws of nature, and natural phenomena (including products of nature) to be such exceptions. Guidance at Background.

If instead the claim is found to be directed to a judicial exception in Step 2A, the analysis continues to Step 2B, which evaluates whether the claimed additional elements amount to significantly more than the recited judicial exception itself (Step 2B). Id. Step 2B includes a consideration of whether the additional element (or combination of elements) is a well-understood, routine, conventional activity. Guidance at II. If an Examiner determines, under Step 2B, that the additional elements do amount to significantly more than the judicial exception, the claim is patent eligible.

To be clear, the July 2024 guidance does not reconfigure the patentability analysis whatsoever with regards to Alice/Mayo, but instead further explains the existing framework, and provides examples of potentially patentable subject matter against that backdrop.

New Subject Matter Eligibility Examples

In the July guidance, the USPTO acknowledges that, “while it is common for claims to AI inventions to involve abstract ideas, USPTO personnel must draw a distinction between a claim that ‘recites’ an abstract idea (and thus requires further eligibility analysis) and one that merely involves, or is based on, an abstract idea.” Guidance at III.A.1. To assist with such a distinction, the text provides examples comprising both patent eligible and ineligible material as evaluated under Step 2A, Prong One. While the provided examples are drawn to mathematical concepts, certain methods of organizing human activity, or mental processes, each of which could, in theory, involve artificial intelligence, the Patent Office does not provide AI-related exemplars under this heading.

The USPTO does, however, reaffirm that an abstract idea must fit into one of the following categories: mathematical concepts, mental processes, and/or “certain methods of organizing human activity,” under which the guidance clarifies that “except in rare circumstances, this grouping should not be expanded beyond the activity within the enumerated sub-groupings of fundamental economic principles or practices, commercial or legal interactions, and managing personal behavior or relationships or interactions between people.” Guidance at III.A.1.B. Regarding AI, the guidance further clarifies that “[t]he mental processes grouping is not without limits, and as such, claim limitations that only encompass AI in a way that cannot practically be performed in the human mind do not fall within this grouping.” Guidance at III.A.1.C (emphasis added). The provided examples indicate the direction the USPTO is headed with regards to examination of patents rejected under 101, but fall short of specifically describing specific AI-related case law.

The Office’s discussion of Step 2A, Prong Two is mildly more fulsome with regards to AI, but lacks any of the specific examples of the preceding prong. Here, the guidance provides clarity regarding differentiation of computer-based claims, and recites that “[a] key point of distinction to be made for AI inventions is between a claim that reflects an improvement to a computer or other technology described in the specification (which is eligible) and a claim in which the additional elements amount to no more than (1) a recitation of the words “apply it” … or are no more than instructions to implement a judicial exception on a computer, or (2) a general linking of the use of a judicial exception to a particular technological environment or field of use (which is ineligible).” Guidance at III.A.2.A. In so stating, the USPTO has reinforced that AI-related improvements remain patent-eligible, and are subject to the same analysis as other computer-related improvements, provided they transcend the mere recitation of “apply it,” or general linking to a technological field of use, and instead provide an improvement to a computer or other technology described in the specification.

Finally, and most crucially, the Guidance provides three hypothetical examples regarding patent eligibility for inventions incorporating AI systems:

  • Example 47 illustrates the application of the eligibility analysis to claims that recite limitations specific to AI, particularly the use of an artificial neural network to identify or detect anomalies. The USPTO’s inclusion of this example (in which claim 1, describing an integrated circuit comprising neurons organized in an array and connected via synaptic circuits, is considered patent eligible) should reinforce the relevance of hardware implementation in AI-related inventions. Claim 3 was similarly found patent eligible, as the claim – drawn to an abstract idea of detecting network intrusion using a neural network model – integrates such an abstract idea into an improvement in the functioning of a computer or technical field, and specifically in  the technical field of network intrusion detection. Exemplary claim 2 was determined ineligible, as merely representing instructions to implement an abstract idea on a computer, with insignificant extra-solution activity.
  • Example 48 illustrates the application of the eligibility analysis to claims that recite AI-based methods of analyzing speech signals and separating desired speech from extraneous or background speech. While the USPTO describes claim 1 – briefly, directed to identifying speech portions using an AI model – as an ineligible abstract idea, it finds claims 2 and 3, incorporating said speech analysis AI model into a processing method and computer system, to be patent eligible. The USPTO recites, for example, that claim 2 “integrates the abstract idea into a practical application,” which enforces its eligibility, because “steps (f) and (g) are directed to creating a new speech signal that no longer contains extraneous speech signals from unwanted source, […and] converting clusters into separate speech waveforms and generating a mixed speech signal from the separate speech waveforms are not insignificant extra-solution activity.”
  • Example 49 illustrates the analysis of method claims reciting an AI model that is designed to assist in personalizing medical treatment to the individual characteristics of a particular patient. Here, while the Office determines that claim 1, which recites using the AI model to recommend a treatment for a patient based on a dataset of patient information and treatment outcomes, is patent ineligible as directed to an abstract idea (mathematical calculation) but allows claim 2, which is directed to a more specific AI model, which the Office specifically describes as “[r]elying on the determination of patient risk to administer Compound X eye drops to glaucoma patients at high risk of PI after microstent implant surgery is therefore a particular treatment for a medical condition such that the claim as a whole integrates the judicial exception into a practical application.” Because the particular treatment integrates the judicial exception into a practical application, it is deemed eligible.  

Guidance at V; Appendix providing July 2024 Subject Matter Eligibility Examples 47-49.

AI Inventorship in Foreign Jurisdictions

Regarding inventorship when it comes to AI, the July 2024 guidance from the USPTO reiterated that the “current statutes … do not provide for recognizing contributions by tools such as AI systems (or other advanced systems) for inventorship purposes, even if those AI systems were instrumental in the creation of the invention.” Further, the guidance emphasizes that “how an invention is developed,” AI or otherwise, “is not relevant to the subject matter eligibility inquiry,” but instead that “the inquiry focuses on the claimed invention itself and whether it is the type of innovation eligible for patenting.” Guidance at IV (emphasis added). Thus, while the action of artificial intelligence may be considered in the invention itself (as described in detail above), it has no rule, under US law, in the inventorship analysis. Indeed, while no country globally has thus allowed an application with a sole listed inventor of AI,[1] the same strict approach has not been universally adopted worldwide. Two examples below demonstrate major patent examining bodies’ approach to AI inventorship.

In the UK, and closely following the approach of the United States, the Supreme Court recently ruled that AI systems cannot be identified as inventors on patent applications. Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49. In that case, Dr. Stephen Thaler’s AI system, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), allegedly invented an invention comprising a food and beverage container. See WO 2020/079499 A1. Dr. Thaler was not listed as an inventor on the application, as he alleged that the DABUS AI system generated the claimed invention autonomously – a factual assertion which was not questioned by the patent examiner at the UK Intellectual Property Office, nor during litigation before the UK courts.

However, the UK Supreme Court nonetheless held that an inventor need be a legal person and natural person, precluding AI from being listed as an inventor. The Court further clarified, citing to the Patents Act of 1977, that an applicant need be either the inventor, under the parameters described above, or a person claiming such an entitlement through the inventor, such as an employer (under certain employment agreements) or successor in interest. Moreover, and further distinguishing the facts, the court held that the DABUS system could not have “devised an invention” (as required for inventorship) under UK law, and reiterated that only a natural person can “devise an invention,” mirroring the US Supreme Court’s approach in precluding non-human inventorship.  

In June 2024, however, the Bundesgerichtshof (the German high court) acquiesced slightly compared to its US and UK counterparts, and held that “the addition of a sufficiently clear designation of the inventor by stating that the inventor has induced a specified artificial intelligence to generate the invention is legally irrelevant and does not justify the rejection of the application.” See machine translation of German opinion, BGH, decision of June 11, 2024. In so holding, the court allowed the following statement of inventorship: “Stephen L. Thaler, PhD, which prompted the artificial intelligence DABUS to generate the invention,” which was raised as Auxiliary Request 3 during appellate proceedings before the German Patent Court.

In allowing such an inventorship designation, the Bundesgerichtshof stated that “it is sufficiently clear from the information [of Auxiliary Request 3] that the applicant names himself as the inventor. The additional – for the reasons explained above legally irrelevant – information on the use of artificial intelligence is easily separable from the designation of the inventor and can be disregarded for the collection and processing of the data.” Id. The German court clarified that “the fact that an artificial intelligence system has made a substantial contribution to the discovery of a technical teaching does not contradict the assumption that there is at least one natural person who is to be regarded as the inventor on the basis of his contribution. Against this background, it is possible and reasonable for the applicant to name (at least) one inventor even if, in his view, an artificial intelligence system has made the main contribution.” Id. Thus, while the Bundesgerichtshof did not go as far as to allow an AI system to be deemed an inventor, the court did allow mention of a natural person’s prompt of an artificial intelligence to generate the invention, breaking with global precedent in this regard.

Dr. Thaler’s legal cases are supported by the Artificial Inventor Project, which handles various “pro bono legal test cases” designed to “promote dialogue about the social, economic, and legal impact of frontier technologies such as AI and to generate stakeholder guidance on the protectability of AI-generated output,” and is led by Professor Ryan Abbott of the University of Surrey.

Conclusion

While the majority of the USPTO’s July 17, 2024, guidance is reiteration of previous law and policy, the three new examples (47-49) provided seem to show the direction of the agency’s stance on AI-related invention, wherein specific hardware implementations and/or technological integrations of AI models – rather than mere application of AI to existing mental processes – weighs in favor of patent eligibility. Thus, the law surrounding AI-incorporating inventions has been clarified by this guidance, and will undoubtedly continue to progress in view of it.

Regarding inventorship, the USPTO’s guidance provided no change to law “that only a natural person can be an inventor, so AI cannot be.” Thaler 43 F.4th  1213 (Fed. Cir. 2022), cert denied, 143 S. Ct. 1783 (2023), but recent global cases – including the Bundesgerichtshof’s holding allowing notation of a human inventor prompting an AI in an inventorship designation – keep the issue at the forefront of patent law.


[1] The “Patent” page on the Artificial Inventor website does list South African patent number ZA2021/03242 as having issued. The authors note that the South African patent office does not conduct substantive examination, and instead only reviews an examination for formalities.