Defining the "Something More": How In re Desjardins and the "SMED" Memo Reshape AI Eligibility

Written by Ben Esplin

For over a decade, the patent eligibility of software and artificial intelligence has been haunted by the vague requirement of Alice Corp.: to demonstrate that a claim contains "something more" than an abstract idea. On December 4, 2025, USPTO Director John A. Squires provided the definitive answer to what that "something more" actually looks like, codifying the recent decision in In re Desjardins as the binding framework for the future of AI prosecution.

The Desjardins Framework

The Memorandum designates In re Desjardins (Appeal No. 2024-000567) not merely as a helpful case, but as the precedential "framework" for evaluating eligibility in new and emerging technologies. In Desjardins, the Board vacated a Section 101 rejection of claims directed to a machine learning model designed to learn multiple tasks without suffering from "catastrophic forgetting"—a common failure mode where new training overwrites previous knowledge.

The Director’s memo explains that the Desjardins claims were eligible because they did not merely use a computer to perform a calculation; they improved the machine learning model itself. The decision specifically credited the claims for reducing storage requirements and lowering system complexity. By designating this decision as precedential, Director Squires has established that improvements in computational performance, learning, storage, and data structures are patent-eligible technological advancements.

Crucially, the Director outlines how Desjardins instructs Examiners to find the "something more." He notes that especially with AI, the "something more" is found by evaluating how the invention is applied, how it operates, and what it accomplishes. It is no longer sufficient for an Examiner to look only at the result; they must look at whether the claimed system changes the architecture itself—specifically "how information flows, not just what it does." If the claim alters the flow of information to improve the tool's functioning, the Desjardins framework dictates that the eligibility requirement is satisfied.

Moving from Argument to Evidence

While Desjardins provides the legal theory, the USPTO’s accompanying guidance on Subject Matter Eligibility Declarations (SMEDs) provides the tactical roadmap for proving it. The guidance makes clear that when attorney argument regarding the "something more" is insufficient, applicants should shift to submitting factual declarations that bridge the gap between the abstract and the practical.

For example, in the context of distributed computing, Examiners frequently allege that claim limitations involving network monitoring fall within the "mental process" grouping of abstract ideas. The new guidance suggests that an applicant can overcome this by submitting expert testimony explaining that the specific, distributed nature of the process—such as analyzing packets across multiple machines simultaneously—is physically impossible for a human mind to perform. This converts the rejection from a philosophical debate into a factual error regarding the state of the art.

Similarly, in cases involving neural networks, a specification might assert that a new architecture is "better" than the prior art, only for the Examiner to dismiss the claim as a mathematical concept. The guidance explicitly validates the use of a declaration to submit objective comparative testing data. If an expert can demonstrate that the claimed architecture achieves superior regression performance compared to the prior art systems identified in the specification, that factual evidence establishes the "improvement to technology" required for eligibility.

The same strategy applies to inventions that automate previously subjective tasks, such as computer animation. An Examiner might reject a claim as merely automating a fundamental human activity. However, a declaration can provide evidence that the specific rules recited in the claim enable the automation of a process that, prior to the invention, could only be performed subjectively by humans. By proving via testimony that the "rules" in the claim solve a problem that previously required manual intervention, the applicant effectively proves the invention is a technical solution rather than an abstract idea.

The Takeaway

The Desjardins framework and the SMED guidance signal a new phase in patent prosecution. We are moving away from abstract inquiries into "inventive concepts" and toward a concrete analysis of system architecture and information flow. By understanding the Desjardins criteria and supporting them with the evidentiary record demanded by the new guidance, applicants can secure robust protection for the next generation of AI innovation.