USTPO AI Examination Update Provides New Example Cases Analyzing Subject Matter Eligibility Under §101 | BakerHostetler
In a July 16 press release,[1] The U.S. Patent and Trademark Office (USPTO) announced that it issued a guidance update[2] on “patent subject matter eligibility to address innovation in critical emerging technologies including artificial intelligence (AI).” According to the press release, this latest update (which went into effect on July 17) “builds on previous guidance by providing further clarity and consistency to how the USPTO and applicants should evaluate subject matter eligibility of claims in patent applications and patents involving inventions related to AI technology. The guidance update also announces three new examples of how to apply this guidance throughout a wide range of technologies.”
The guidance update relabels Alice/Mayo Step 1 and Step 2 as Step 2A and Step 2B. The flowchart for Step 2A is provided as follows:
With respect to AI- and software-based inventions, Prong One of this step focuses on the traditional Alice/Mayo Step 1 of whether the claim recites an “abstract idea.” Categories of abstract ideas include mathematical concepts, certain methods of organizing human activity (e.g., fundamental economic principles; commercial or legal interactions; managing personal behavior, relationships or interactions between people) and mental processes. With respect to mental processes and AI, the guidance update has the following useful passage: “[C]laims do not recite a mental process when they contain limitations that cannot practically be performed in the human mind, for instance when the human mind is not equipped to perform the claim limitations. … [C]laim limitations that only encompass AI in a way that cannot practically be performed in the human mind do not fall within this grouping.”
If not directed to an abstract idea, the claim is directed to eligible subject matter; but if it is directed to an abstract idea, the test then goes to Prong Two. Prong Two asks whether the claim recites “additional elements that integrate the [abstract idea] into a practical application.” If it does, then the claim qualifies as eligible subject matter.
The “practical application” test is based on “judicial considerations” from prior Federal Circuit cases:
The guidance update itself provides several interesting examples for most of the above scenarios.
If the claim fails Prong Two, the test goes to Step 2B (Alice/Mayo Step 2), which is whether the claim recites “significantly more” than the abstract idea. The guidance update does not go into as much detail on this step, essentially explaining that “Step 2B includes a consideration of whether the additional element (or combination of elements) is a well-understood, routine, conventional activity.” The guidance update does note that the examiner must provide factual support if this is the conclusion.
The guidance update also introduces three new helpful examples[3] for the AI subject matter eligibility test: Example 47, “Anomaly Detection”; Example 48, “Speech Separation”; and Example 49, “Fibrosis Treatment.” The examples are interesting in that each case has some claims that are found ineligible and some that are found eligible. It is enlightening to see the difference between those found ineligible and those found eligible.
Example 47 concerns claims for using AI to detect anomalies in data that deviate from expected data or from a general pattern and analyzes three separate claims.
Claim 1 is directed to an application-specific integrated circuit (ASIC) that includes a plurality of neurons organized in an array and a plurality of synaptic circuits interconnecting the neurons. Claim 1 is found eligible because it does not recite an abstract idea in Step 2A, Prong One.
Claim 2 is directed to a method of using an artificial neural network (ANN) that generally includes steps of (a) receiving training data, (b) discretizing the training data, (c) training the ANN, (d) detecting anomalies using the ANN, (d) analyzing the detected anomalies and (e) outputting the result of the analysis. Claim 2 is found ineligible essentially because it recites merely a combination of performing mathematical calculations and using mental choices for analyzing and outputting results.
Claim 3 is substantially more specific and recites in detail how the ANN can be used to detect malicious network packets from detected anomalies in network traffic and then take specific action upon such detection. Claim 3 is found eligible because it integrates the abstract idea into a practical application (Step 2A, Prong Two). Specifically, the claim improves the functioning of a computer in the technical field of network intrusion detection.
Example 48 concerns claims for using AI to analyze speech signals and separate desired speech from extraneous or background speech.
Claim 1 is directed to a speech separation method that generally includes steps of (a) receiving mixed speech signals from different sources, (b) converting the mixed speech signal into a spectrogram of time-frequency domain using short-time Fourier transform and so forth, and (c) using a deep neural network (DNN) to determine embedding vectors from the mixed speech signal. Unsurprisingly, Claim 1 is found ineligible because it merely combines multiple abstract ideas in the form of mathematical concepts, does not claim a solution to a technical problem and provides insignificant extra-solution activity.
Claim 2 depends from Claim 1 and recites specific details for the method that are described in the specification as providing a technical solution to a stated technical problem. While Claim 2 essentially further details the mathematical concepts found ineligible in Claim 1, Claim 2 is considered eligible (Step 2A, Prong Two) because it “reflects the improvement discussed in the disclosure by reciting details of how the DNN aids in the cluster assignments to correspond to the sources identified in the mixed speech signal.”
Claim 3 is directed to nontransitory computer memory that includes computer-executable instructions to cause one or more processors to generally perform the steps of (a) receiving a mixed speech signal … at a DNN trained on source separation, (b) using the DNN to convert a time-frequency representation of the mixed speech signal into embeddings in a feature space, (c) clustering the embeddings using a specified clustering algorithm, (d) applying binary masks to the clusters to obtain masked clusters, (e) converting the masked clusters into a time domain to obtain separated speech signals and (f) extracting spectral features from the separated speech signals and generating a sequence of words to produce a transcript of the speech signal. Claim 3 is considered eligible (Step 2A, Prong Two) because the additional steps (e) and (f) “integrate the abstract idea recited in steps (b), (c), and (d) into a practical application of speech-to-text conversion. … Here, the claim reflects [the] technical improvements discussed in the disclosure by reciting details of how the DNN trained on source separation aids [in solving the stated technical problems]. … Accordingly, the claim is directed to an improvement to existing speech-to-text technology.”
Example 49 concerns claims directed to using AI to determine an appropriate treatment for a glaucoma patient at high risk of postimplantation inflammation (PI) after microstent implant surgery. Example 49 also concerns “a new anti-fibrotic drug, Compound X,” developed by the applicant for such treatments.
Claim 1 is directed to a postsurgical fibrosis treatment method that generally includes the steps of (a) collecting a genotyping sample from the patient to provide a genotype dataset, (b) identifying that the patient is at high risk of PI based on a risk score that is generated from single nucleotide polymorphisms in the dataset by an AI model that uses multiplication to weight corresponding alleles in the dataset and sum the weighted values to produce a score, and (c) administering an appropriate treatment to the high-risk patient. Claim 1 is found ineligible as being merely directed to a combination of abstract ideas and further abstract improvements to those ideas. While the disclosure states that step (b) improves upon the base polygenic risk score model by determining the risk score and providing classification in less time, “there is no improvement to the functioning of a computer nor to any other technology. At best, the claimed combination amounts to an improvement of the abstract idea of determining patient risk rather than to any technology.”
Claim 2 depends from Claim 1 and simply adds “wherein the appropriate treatment is Compound X eye drops.” Claim 2 is considered eligible (Step 2A, Prong Two) as providing “a particular treatment for a medical condition such that the claim as a whole integrates the judicial exception into a practical application.” The analysis spends a substantial focus on the recognition that Compound X is identified in the disclosure as a new treatment and not “any common anti-fibrotic treatment,” stating that administering “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.” In other words, it appears the analysis considers Claim 2 eligible because Compound X is an improvement to the technical field of glaucoma treatments and reciting any common glaucoma treatment in Claim 2 would have resulted in an ineligible claim.
Takeaways
The guidance update and new subject matter eligibility examples 47-49 provide useful insights into how to draft patent applications and claims with an eye toward eligibility, primarily under the Step 2A, Prong Two, “practical application” test. It is notable that each of the examples relies on the specification’s disclosure to determine whether any of the claim elements provide a technical solution to a technical problem. These further drive home the importance of a robust patent specification that includes significant details as to the implementation, customization and training of the AI/machine learning (ML) models. Practitioners should be sure that the specification explains in detail how the AI/ML technology is being applied to a practical application or integrated with another system or apparatus. Practitioners should also be sure that the specification explains technical problems confronted in the development and how those technical problems were overcome by the invention. Finally, practitioners should be sure that at least some of the claims expressly recite the specific technical solutions as stated in the specification as resolving the stated technical problems. It could be that these claim limitations are the ones that survive the eligibility test.
[1] The link to the press release is
[2] The link to the guidance update is
[3] The link to the new examples is
[View source.]
