Source: Crowell & Moring

Crowell & Moring analyzes the landmark decision Ex parte Desjardins, which the USPTO designated as precedential on November 4, 2025. The Appeals Review Panel (ARP) overruled a previous decision by the Patent Trial and Appeal Board, holding that claims directed to training a machine learning model are patent-eligible under 35 U.S.C. § 101. The case involved a computer-implemented method for sequentially training models while preventing "catastrophic forgetting." While the board initially viewed the mathematical calculations as abstract ideas, the ARP found that the specific adjustment of parameters to optimize performance across multiple tasks constituted a technological improvement to how the machine learning model itself operates.

This decision marks a significant shift in favor of AI and software eligibility, as the ARP warned that categorically excluding AI innovations from protection could jeopardize America’s technological leadership. Following this ruling, the USPTO issued updates to the Manual of Patent Examining Procedure (MPEP) on December 5, 2025, instructing examiners to evaluate specifications for non-conclusory disclosures of technological improvements. Examiners are now cautioned against overgeneralizing algorithmic steps as "mental processes" or dismissing limitations as "generic computer components." This development provides patent practitioners with concrete tools to argue for the eligibility of machine learning inventions that enhance computer functionality through practical applications.