Summary of the Appeals Review Panel Decision in Ex Parte Desjardins (September 26, 2025)

Summary of the Appeals Review Panel Decision in Ex Parte Desjardins (September 26, 2025)

In late September 20225, in Ex Parte Desjardins, the Director of the United States Patent and Trademark Office (USPTO), John Squires, convened an Appeals Review Panel (ARP) that vacated a § 101 rejection concerning claims directed to methods of training machine learning models. Importantly, the vacatur of a PTAB decision by the Director is a rare event, particularly when issued sua sponte through an Appeals Review Panel. The use of this extraordinary procedural mechanism signals the USPTO’s heightened awareness of the implications of § 101 jurisprudence on AI and software-related inventions and suggests a potential policy shift toward a more permissive posture in such technologies.  This article discusses the ARP’s decision in Ex Parte Desjardins.

The ARP determined that the claims were not directed to an abstract idea but instead integrated the abstract idea into a practical application, thereby satisfying subject matter eligibility under 35 U.S.C. § 101.  The patent application, filed by DeepMind Technologies Limited, discloses a method for training a machine learning model on sequential tasks.

A key limitation of the independent claims involved “computing… an approximation of a posterior distribution over possible values of the plurality of parameters” and using this to preserve performance on a first task while training on a second task. The Patent Trial and Appeal Board (PTAB) had earlier affirmed a § 103 rejection and entered a new ground of rejection under § 101, categorizing the claims as reciting a mathematical concept and, therefore, as an abstract idea.

On rehearing, the ARP focused on Alice Step 2A, Prong Two, evaluating whether the claims integrated the abstract idea into a practical application. The panel found that the claims reflected specific improvements in the operation of the machine learning model itself – such as enabling continual learning across tasks while reducing storage requirements and system complexity. These improvements were both described in the specification and supported by the claim language.

The ARP criticized the PTAB’s reasoning for treating machine learning broadly as an unpatentable “algorithm” and for dismissing the additional elements as merely generic computer components. The decision emphasized that such generalized reasoning risks excluding artificial intelligence innovations from patent protection, potentially undermining U.S. leadership in this critical technological area. The panel reaffirmed that §§ 102, 103, and 112 – not § 101 – are the appropriate statutory tools for determining the proper scope of patent protection.

Based on my own experience in interacting with USPTO examiners regarding software patent applications, this decision aligns with what one examiner recently communicated to me: that the USPTO may be becoming more permissive toward allowing software-related patent applications when they are drafted carefully. Specifically, the examiner emphasized the importance of clearly articulating technical improvements and practical applications in both the claims and the specification. This anecdotal insight is consistent with the reasoning and tone of the Desjardins decision and suggests a meaningful shift in the USPTO’s approach to software and AI-related subject matter eligibility.

Ultimately, the ARP vacated the § 101 rejection, while leaving the § 103 rejection intact. This decision signals a potentially more favorable path forward for AI and software-related inventions, provided applicants can demonstrate how their innovations improve underlying technology in a concrete and practical way. Click HERE to learn more about Derek Fahey, Esq., patent attorney and author of this article.