Section 101 Patent Eligibility Update: The Tech-Architecture Divide

Recent actions by the United States Patent and Trademark Office (USPTO) and the U.S. Court of Appeals for the Federal Circuit have provided new clarity for software, artificial intelligence (AI) and machine learning (ML) claims. Under this new guidance, claims directed to generic data manipulation are expected to be routinely rejected, whereas those reciting specific technological improvements are expected to be patent-eligible. Understanding this divide is essential to navigating Section 101 rejections during prosecution under the Alice/Mayo framework.

The Pro-AI Wave: Ex parte Desjardins and Ex parte Carmody

The USPTO has updated its examination guidance to provide clearer parameters for evaluating the subject matter eligibility of AI and ML subject matter in view of recent judicial precedent.

• The Desjardins Standard: In late 2025, the USPTO designated the Appeals Review Panel decision Ex parte Desjardins as precedential. The holding establishes that claims directed to training or optimizing a machine learning model are patent-eligible under Step 2A, Prong Two if they improve the internal operation of the model itself. The USPTO codified this by expanding the Manual of Patent Examining Procedure (MPEP) to credit benefits like “reduced storage,” “reduced system complexity” and avoiding “catastrophic forgetting” as eligible technical improvements, specifically at MPEP § 2106.04(d), Subsection III.
• The “Whole Workflow” Expansion: In several instances, examiners have attempted to assert that Desjardins only applies to pure, isolated machine learning models (vs. claims that include additional elements beyond the machine learning elements). However, the Patent Trial and Appeal Board (PTAB) explicitly shut down this “model-only” restriction in Ex parte Carmody (PTAB Dec. 30, 2025). Carmody confirms that integrating an AI or ML engine into a broader business workflow or downstream application is not a bar to eligibility, provided the claim evaluated “as a whole” achieves an underlying technological optimization.

The High-Generality Trap: In re Brian McFadden

While Desjardins opened a clear path forward, the Federal Circuit defined a strict boundary line in In re Brian McFadden (Fed. Cir. April 2026).

In McFadden, the court affirmed the rejection of a patent application directed to an information exchange apparatus, noting that the applicant’s mathematical algorithms merely calculated information distribution differences at a high level of generality. The Federal Circuit held that:

“These instructions for how a standard computer is to manipulate, transform, and compare data fare no better in the abstract idea analysis than generic software or computing components.”

McFadden reaffirms the rule that enhancing abstract calculations on standard hardware, without a specific technological improvement defining how the implementation physically or logically alters computer operation, remains an ineligible abstract idea.

Actionable Key Takeaways for Prosecution

To safely navigate the current § 101 landscape, applicants should consider modifying their drafting and response protocols to leverage the following tools:

If all else fails, consider use of Subject Matter Eligibility Declarations (SMEDs): In light of recent USPTO Director memoranda, applicants may benefit from utilizing SMEDs under Rule 132. These expert-backed declarations permit applicants to submit factual, objective evidence asserting that an ordered combination of elements is structurally unconventional or that a claimed AI step cannot practically be executed in the human mind. However, practitioners should approach SMEDs with caution, as creating a rigid, formal record regarding the state of the prior art can potentially restrict claim scope during any subsequent litigation.

Incorporate the “Desjardins Formula”: Ensure that the specification details a non-conclusory technological problem and solution, and that the independent claims explicitly mirror that technical detail (e.g., reciting specific data structures, recursive loops or dynamic parameter updates).

Weaponize the Preponderance Standard: Remind examiners that under MPEP §§ 706 and 2106, and current USPTO guidance, a § 101 rejection requires a “more likely than not” (>50%) threshold of proof to establish a prima facie case. Rejections rooted in examiner “uncertainty” or speculative assertions fail to satisfy this initial evidentiary burden and must be withdrawn as procedurally and substantively improper.

---

This document is intended to provide you with general information regarding recent changes at the USPTO and Federal Circuit. The contents of this document are not intended to provide specific legal advice. If you have any questions about the contents of this document or if you need legal advice as to an issue, please contact the attorneys listed or your regular Brownstein Hyatt Farber Schreck, LLP attorney. This communication may be considered advertising in some jurisdictions.