Patent claims must be directed to one of four statutory categories and must not be directed to a judicial exception unless the claim as a whole includes additional limitations amounting to significantly more than the exception. The judicial exceptions identified by the Manual of Patent Examining Procedure (MPEP) include abstract ideas, laws of nature, and natural phenomena. With the rise of artificial intelligence (AI) and machine learning (ML) inventions, the USPTO is continually publishing guidelines and reminders for how to evaluate claims for subject matter eligibility.

In the most recent USPTO memorandum, published on August 4, 2025, Deputy Commissioner for Patents Charles Kim issued a memorandum to Examiners in three technology centers that commonly handle patent applications directed to artificial intelligence (AI) and machine learning (ML), which are often rejected for lack of patent eligibility under 35 U.S.C. § 101. This memorandum provided reminders and clarifications for how Examiners should apply eligibility standards to these types of inventions. Some key takeaways described below can help practitioners and companies update their patent strategies.

The memorandum begins by reminding examiners that a broadest reasonable interpretation of the pending claims must be established prior to analyzing the claims for subject matter eligibility. The memorandum then begins a discussion of the mental process sub-grouping of abstract ideas. The memorandum specifically warns examiners to not expand the mental process sub-category in a manner that captures claim limitations that cannot practically be performed in the human mind. In particular, claim limitations that involve AI in a way that cannot be practically performed in the human mind do not fall within the mental process sub-grouping of ineligible “abstract ideas.” The memorandum references Example 39 and Example 47 of the Subject Matter Eligibility Examples previously published by the USPTO to provide further clarity with respect to the mental process sub-group.

Example 39 analyzes a claim directed to a computer-implemented method of training a neural network for facial detection. The claim includes steps of collecting digital facial images, applying one or more transformations to the digital facial image, creating two training sets, and training a neural network using each training set. This example notes that this claim does not recite a judicial exception to patent eligibility because it does not recite any mathematical concepts, mental processes, or methods of organizing human activity. In particular, the steps cannot practically be performed in the human mind. The memorandum notes that although this claim uses broad language of “training the neural network,” there are no specific mathematical concepts that are described and thus, this claim language cannot be interpreted as either a mathematical concept or a mental process.

Example 47, claim 2, is directed to a method of using an artificial neural network (ANN). This claim includes steps of receiving continuous training data, discretizing the training data to generate input data, training the ANN by a backpropagation algorithm and a gradient descent algorithm, detecting one or more anomalies in a data set with the trained ANN, analyzing the detected anomalies to generate anomaly data, and outputting the anomaly data. This claim was deemed to recite a judicial exception because the steps are all either mental processes or mathematical calculations. The memorandum notes that this claim recites a judicial exception in part because the claim requires specific mathematical calculations of a backpropagation algorithm and a gradient descent algorithm.

The memorandum then provides clarity on the analysis that is to be performed by an Examiner if a claim is determined to recite a judicial exception. An Examiner is required to determine whether the claim, as a whole, is integrates the alleged judicial exception into a practical application. If an additional limitation or combination of limitation in the claim meaningfully limits the judicial exception, the claim may be rendered eligible. Importantly, the Examiners should consult the specification to determine whether the invention improves technology or a technological field and should determine whether the claims reflect the disclosed improvement. The claims should reflect this improvement but do not have to explicitly recite the improvement in order to be deemed subject matter eligible. Examiners are also reminded that they should determine whether the claim elements provide more than instructions to implement an abstract idea on a computer.

Further, the memorandum notes that a claim should only be rejected if it is more likely than not that the claim is subject matter ineligible. Thus, unless the Examiner believes there is more than a 50% chance that the claim is ineligible, a rejection under 35 U.S.C. § 101 should not be issued.

While this memorandum does not provide any new practices or procedures for Examiners to follow, it is a good overview of procedure that should be followed by Examiners when rejecting a claim for lack of subject matter eligibility, and can help guide companies as they continuously evaluate their patent portfolio management practices. Further, this memorandum serves as a reminder to patent practitioners to provide a discussion of technical improvements and practical application of inventions within the specification and to ensure that those improvements are reflected in the claims to be examined. Finally, although the USPTO guidance is not binding on courts, district court judges often cite such guidance as persuasive reasoning as to how eligibility issues should be addressed, so this memorandum can also be helpful in pending litigation where patent eligibility is challenged.