L'USPTO clarifie les critères d'examen d'éligibilité pour les innovations logicielles
On August 4, 2025, the U.S. Patent and Trademark Office (USPTO) released a new memorandum to patent examiners in Technology Centers 2100, 2600, and 3600, providing targeted reminders on evaluating subject matter eligibility (SME) under 35 U.S.C. § 101 for software-related inventions, including Artificial Intelligence (AI) and Machine Learning (ML) technologies. The guidance, issued by Deputy Commissioner for Patents Charles Kim, reinforces existing USPTO practice and clarifies common issues in applying the agency’s Step 2A analysis of the Alice/Mayo framework.
While the memorandum does not establish new policy, it provides additional emphasis on four key examination topics:
- Reliance on the “mental process” grouping of abstract ideas;
- Distinguishing between claims that recite a judicial exception and those that merely involve one;
- Analyzing the claim “as a whole” under Step 2A Prong Two; and
- Differentiating claims that improve technology from those that merely “apply” an abstract idea using generic computing resources.
The memorandum also includes guidance on when an SME rejection under 35 U.S.C. § 101 should and should not be made, reinforcing that rejections should only be issued when ineligibility is more likely than not.
It is anticipated that the memorandum will provide more avenues for overcoming 35 U.S.C. § 101 rejections by offering clearer guidelines and emphasizing the importance of a thorough analysis. By focusing on the distinction between claims that recite a judicial exception and those that merely involve one, examiners are better equipped to identify eligible subject matter. Additionally, the memorandum’s emphasis on analyzing claims “as a whole” under Step 2A Prong Two, and differentiating claims that improve technology from those that merely apply an abstract idea using generic computing resources helps ensure that innovative software-related inventions are not unduly rejected. This guidance ultimately supports inventors in navigating the complexities of subject matter eligibility, fostering a more predictable and transparent examination process.
- Step 2A Prong One: Judicial Exceptions and the Mental Process Grouping
The USPTO’s subject matter eligibility analysis recognizes the three previously established categories of abstract ideas: (1) mathematical concepts, (2) certain methods of organizing human activity, and (3) mental processes. The memorandum reiterates that a “mental process” covers concepts that can be performed entirely in the human mind or with pen and paper, such as making observations, evaluations, judgments, or forming opinions.
However, the memorandum warns examiners not to expand this grouping to capture claim limitations that cannot practically be performed in the human mind. For example, limitations involving AI-based processing that go beyond human mental capabilities fall outside this category. The memorandum references the July 2024 AI-SME Update, which provides examples where AI-specific hardware or processes are not considered mental processes (e.g., a hardware-based RFID serial number data structure as in ADASA Inc. v. Avery Dennison Corp., 55 F.4th 900 (Fed. Cir. 2022)).
- Distinguishing “Recites” from “Involves” a Judicial Exception
The memorandum reiterates the need to distinguish claims that merely involve an exception from those that actually recite it. If a claim limitation does not set forth or describe the exception, for example, “training a neural network using a first training set” (Example 39), it may involve abstract concepts but does not itself recite an abstract idea, and thus does not trigger further eligibility analysis under Step 2A.
In contrast, if the claim explicitly names specific mathematical algorithms, such as “training an ANN using a backpropagation algorithm and gradient descent algorithm” (Example 47), it recites a mathematical concept and requires further analysis.
- Step 2A Prong Two: Practical Application and Improvements
Under Prong Two, examiners must evaluate whether the claim integrates any recited judicial exception into a practical application. This requires an analysis of the claim as a whole, taking into account how all limitations interact, rather than isolating them in a vacuum.
The memorandum highlights two major considerations that often overlap:
- Improvements to technology or a technical field – Claims that present a specific, technical solution to a technical problem can satisfy Prong Two. The memorandum emphasizes that “the specification does not need to explicitly set forth the improvement, but it must describe the improvement in a way apparent to a person of ordinary skill in the art, even if the claim itself does not explicitly recite the improvement.”
- Avoiding “Apply It” claims – Merely instructing a computer to carry out an abstract idea without specific technological improvements will not suffice. Examiners are advised not to oversimplify claims when applying this test and to consider whether the claim provides a particularized implementation rather than a generic application.
The memo notes that Recentive Analytics, Inc. v. Fox Corp. (Fed. Cir. 2025) illustrates the insufficiency of steps incidental to automating an abstract idea, while USPTO Example 47, claim 3, demonstrates a claim that improves the technical field of network intrusion detection.
- When to Issue a § 101 Rejection
Examiners are reminded that eligibility rejections under 35 U.S.C. § 101 should not be issued simply due to uncertainty. A rejection is appropriate only if it is more likely than not that a claim is ineligible, applying the “preponderance of the evidence” standard. This “close call” guidance is intended to prevent over-rejection in borderline cases and to focus examination on complete, compact prosecution addressing all statutory requirements (35 U.S.C. §§ 101, 102, 103, 112) in the first Office action.
Impacts on Businesses
For innovators in software, AI, and ML, this memorandum signals the USPTO’s renewed emphasis on eligibility determinations regarding software-related inventions. The memorandum creates a clearer path to patenting software-related inventions by narrowing overuse of the “mental process” category and reaffirming Example 39, meaning inventions that involve software, and particularly AI or ML, do not themselves trigger an abstract-idea finding at Prong One. Additionally, the memorandum allows software-related inventions to pass Step 2A more predictably based on technical improvements provided by the invention, even if the claims and the specification do not explicitly state the technical improvement, as long as “the improvement would be apparent to one of ordinary skill in the art.” In prosecution, the >50% “close-call” standard provides a concrete lever to push back on tentative 35 U.S.C. § 101 rejections and keep the focus on §§ 102/103/112. As a result, applicants can expect less eligibility risk, a shorter path to allowance, and greater predictability in budgeting and portfolio planning.
