Bot or Not? – Updated USPTO Guidelines on Patent Eligibility of AI-Related Inventions

One of the requirements for obtaining a patent in the United States is that the invention falls within the eligible categories of subject matter. The requirement is outlined in 35 U.S.C. 101 [1], which states:

“Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.”

At first blush, this would seem to be a fairly permissive definition, permitting the grant of a patent for “anything under the sun that is made by man.”

AI-related inventions are at the forefront of today’s tech revolution, encompassing machine learning and neural networks to natural language processing and intelligent systems. These innovations are transforming industries by enabling new capabilities, boosting efficiency, and solving complex problems that were once thought to be unsolvable.

The requirement stated above has been interpreted over the years to modulate its apparent breadth.  Today, to determine if a computer-related invention is eligible for a patent, the courts often use a two-step framework known as the Alice [2] / Mayo [3] test. This test, established by the U.S. Supreme Court first asks if the claims are directed to something unpatentable, like an abstract idea, natural phenomenon, or law of nature. If they are, the second step determines whether the claim provides something significantly more that turns the abstract idea into a practical, patent-eligible application.

Computer-related inventions, including those involving AI, inevitably face increased scrutiny regarding their eligibility as subject matter, with some applications being rejected for being merely abstract ideas. The United States Patent and Trademark Office (USPTO) recently released updated guidance on patent eligibility for AI and emerging technologies [4].

This guidance aims to clarify subject matter eligibility assessments for USPTO Examiners and help stakeholders, including inventors, understand eligibility for inventions involving AI features and related technologies.

The guidance reiterates the application of the Alice/Mayo test and discusses its application for AI-related inventions. Specifically, it addresses Step 2A (Part 1), which evaluates whether a claim recites an abstract idea, and Step 2A (Part 2), which assesses whether the claim includes additional elements that significantly exceed the abstract idea and provides technical improvements or a practical application. The rationale behind the guidance is to prevent patent owners from gaining monopolistic control over basic scientific or technological principles while providing clarity to applicants as to what may or may not be considered acceptable.

Step 2A (Part 1) states that patent claims directed solely to a judicial exception (e.g., an abstract idea) are ineligible and assesses if the claim includes such an exception. The guidance assists USPTO Examiners in distinguishing whether AI invention claims cover only an abstract idea (ineligible) or use an abstract idea for a practical purpose (eligible). For instance, a claim for an ASIC (application-specific integrated circuit) for an artificial neural network with an array of neurons and synaptic circuits with specific structural elements is not considered an abstract idea. On the other hand, mathematical concepts, methods of organizing human activity, and mental processes are common categories of abstract ideas. For example, a patent claim for collecting and electronically recording a user’s movements and location history is ineligible as it merely manages personal behaviour or interactions. Similarly, a claim for detecting fraud in financial transactions by matching two financial records is ineligible as it covers a fundamental economic principle. The ineligibility of “mental processes” is based on the premise that activities performable in the human mind or with pen and paper are abstract ideas. For example, claims using an algorithm-generated content-based identifier for data management functions, such as controlling access, retrieving, delivering, and marking data for deletion, are ineligible as they cover a multistep mental process performable in the human mind.

Step 2A (Part 2) is triggered when Part 1 finds a judicial exception. Part 2 assesses whether the claim includes elements that significantly exceed the abstract idea, providing technical improvements or a practical application, considering the claim as a whole. Regarding AI inventions, the guidance notes that additional elements merely stating “apply it” do not make an abstract idea eligible. Similarly, generally linking an abstract idea to a technological environment is also ineligible.

Overall, the guidance reiterates that claims, including those involving AI inventions, that improve computer functionality or provide a specific method to achieve an intended outcome are acceptable subject-matter integrations of an abstract idea into a practical application. However, an improvement in the judicial exception or an abstract idea itself is not an improvement in technology. For example, the guidance illustrates that a claim to resolve a haplotype phase based on received allele data using a Hidden Markov Model, even if it yields more haplotype phase predictions, provides improvement to the mathematical process and not the computer technology, and is thus ineligible subject matter. Citing another example, the guidance mentions that claims to a rule-based system for animating lip synchronization and facial expressions of 3D characters, where automation of previously unautomatable tasks was provided, is considered a valid improvement. Similarly, claims to perform error correction and detection encoding with information bits in variable subsets were directed to an eligible improvement in encoding data, relying on irregular repetition and not an abstract idea.

What is more interesting is that the guidance clarifies, for subject matter eligibility analysis under 35 U.S.C. 101, that an invention which was created with the assistance of AI is not a consideration in the application of the Alice/Mayo test. Issues related to AI-assisted inventions were covered in another USPTO guidance issued earlier this year [5].

To assist Examiners when assessing subject-matter eligibility for AI-related inventions, the USPTO has provided three examples [[6]] for AI inventions. The examples provide a subject matter eligibility analysis under 35 U.S.C. 101 of hypothetical claims.

Example 47

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 (ANN) to identify or detect anomalies.

The first exemplary claim was directed to an application-specific integrated circuit (ASIC) for an artificial neural network. The claim for an ASIC for an artificial neural network was determined to be patent-eligible.

The second exemplary claim was directed to a method of using an ANN comprising the steps of: receiving continuous training data, training an ANN; detecting anomalies; and outputting the anomaly data from the trained ANN.

The claim to the method of using an ANN was determined to be patent ineligible. Specifically, the steps recited in the method were all recited as being performed by a generic computer performing generic computer functions. The recitation of a computer to perform the steps of the claim amounts to no more than mere instructions to apply the exception using a generic computer component and does not provide an inventive concept.

The third exemplary claim was directed to a method of using an ANN to detect malicious networks, comprising the following steps: training an ANN; detecting anomalies; detecting a source address in real time; dropping malicious network packets in real time; and blocking future traffic from the source address.

The third exemplary claim directed to the method of using an ANN to detect malicious networks was determined to be patent-eligible. Specifically, this claim improves the functioning of a computer or technical field. The claimed invention reflects this improvement in the technical field of network intrusion detection. Thus, the claim integrates the judicial exception into a practical application and is eligible.

Example 48

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.

The first exemplary claim is directed to a speech separation method. The method comprises: receiving a mixed speech signal, converting the mixed speech signal into a spectrogram, and using a deep neural network to determine embedding vectors.

It was determined that the first exemplary claim is ineligible as it merely provides instructions to apply an exception and insignificant extra-solution activity and therefore does not provide an inventive concept.

The second exemplary claim is a claim dependent on the first claim. The second claim relies on the method steps described by the first exemplary claim and adds the further steps of partitioning the embedding vectors, applying binary masks to create masked clusters, synthesizing speech waveforms; combining the speech waveforms to generate a mixed speech signal; and transmitting the mixed speech signal for storage to a remote location.

The second exemplary claim was determined to be patent-eligible because the claim is directed to an improvement to existing computer technology or technology of speech separation, and the claim integrates the abstract idea into a practical application.

The third exemplary claim is directed to a non-transitory computer-readable storage medium having computer-executable instructions stored thereon. This claim is directed to an improvement to existing speech-to-text technology, and the claim integrates the abstract idea recited in steps (b), (c), and (d) into a practical application of speech-to-text conversion of a speech signal. Thus, the claim integrates the judicial exception into a practical application and is therefore patent-eligible.

Example 49

Example 49 shows the analysis of method claims reciting an AI model designed to help personalize medical treatment to the individual characteristics of a particular patient.

The first exemplary claim is directed to a post-surgical fibrosis treatment method comprising the following steps: collecting and genotyping a sample to provide a genotype dataset; identifying the patient as at high risk using an AI model; and administering an appropriate treatment to the patient.

It was determined that the first exemplary claim is ineligible as it merely provides instructions to apply an exception and insignificant extra-solution activity and therefore does not provide an inventive concept.

The second exemplary claim is a claim dependent on the first exemplary claim. The second claim relies on the method steps described by the first exemplary claim and further clarifies that the appropriate treatment is eye drops. The second exemplary claim is eligible as the claim is not directed to a judicial exception.

The guidance, with examples of AI inventions, emphasizes that claims including AI features directed at improving computer functionality or achieving a specific practical application are likely to be considered eligible subject matter. Subject matter jurisprudence continues to evolve, so applicants are advised to stay updated on developments and discuss potential subject matter eligibility issues with a qualified patent professional.

 

If you require legal advice or more information regarding Patent Eligibility of AI-Related Inventions, please contact a member of Bereskin & Parr’s AI practice group.