AI Patent Eligibility Dramatically Expanded at USPTO
The USPTO's September 26, 2025 decision in Ex parte Desjardins represents the most significant expansion of AI patent eligibility since Alice created the modern Section 101 framework in 2014. Now designated as precedential, this Appeals Review Panel (ARP) decision. personally authored by newly confirmed Director John Squires, fundamentally recalibrates how the USPTO examines AI and machine learning patent applications.
The Decision's Core Holdings
Director Squires' decision establishes that improvements to how machine learning models operate constitute patent-eligible technological advances when claims reflect specific technical benefits. The case involved a Google/DeepMind application addressing "catastrophic forgetting" in continual learning systems, where ML models lose knowledge of previous tasks when trained on new ones. The ARP found the claims patent-eligible because they improved the ML model's operation by preserving performance across sequential tasks while reducing storage requirements.
Most critically, the decision warns against evaluating AI claims at excessive levels of generality. Director Squires criticized the PTAB panel's "overbroad" and "troubling" approach that "essentially equated any machine learning with an unpatentable 'algorithm' and the remaining additional elements as 'generic computer components,' without adequate explanation."
Practical Framework for AI Patent Eligibility
The Desjardins framework distinguishes patent-eligible AI innovations through several key principles:
Patent-Eligible Claims Must Show:
Improvements to how the machine learning model itself operates
Technical solutions to recognized problems (e.g., catastrophic forgetting)
Specific architectural or training improvements
Concrete measurable benefits like storage reduction or efficiency gains
Insufficient for Eligibility:
Merely applying generic ML to new data environments
Field-of-use limitations without model improvements
Broad invocations of "any suitable machine learning technique"
Claims focused on application domains rather than model functionality
Navigating the USPTO-Federal Circuit Divide
While Desjardins creates favorable examination conditions at the USPTO, tension remains with the Federal Circuit's April 2025 Recentive Analytics decision. That ruling held that "patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible."
This divergence creates strategic complexity. Applications will receive more favorable treatment at the USPTO under binding Desjardins precedent, but issued patents may face stricter Section 101 challenges in litigation under Recentive. We recommend dual claiming strategies: broader independent claims leveraging the USPTO's framework, paired with narrower dependent claims containing implementation details for litigation defense.
Specification Drafting in the Post-Desjardins Era
Success under Desjardins requires specifications that articulate genuine technological advancement rather than mere application of AI to new problems. Following the Enfish precedent that the decision heavily relies upon, specifications should establish clear technical narratives:
Problem-Solution Architecture: Frame innovations as solutions to concrete technical challenges in AI systems themselves. Rather than describing business problems solved by AI, focus on computational inefficiencies, architectural limitations, or performance bottlenecks that your innovation addresses.
Technical Improvement Documentation: Quantify improvements wherever possible—whether through computational complexity reduction, memory optimization, accuracy enhancements, or processing speed gains. These metrics transform abstract algorithmic concepts into concrete technological advances.
Implementation Depth: While avoiding unnecessary limitations, provide sufficient technical detail to demonstrate that the innovation extends beyond high-level functional claiming. Architecture diagrams, data flow illustrations, and comparative performance analyses strengthen eligibility arguments without constraining claim scope.
Alternative Embodiments: Include multiple implementation pathways to demonstrate that you're claiming a technological solution rather than an abstract result. This approach supports broader claims while providing fallback positions for prosecution or litigation.
Strategic Recommendations for Patent Portfolios
The Desjardins framework creates immediate opportunities requiring prompt action:
Accelerate Filing Timelines: File AI patent applications while the favorable examination climate persists, particularly for innovations involving model architectures, training improvements, or efficiency enhancements.
Leverage Continuation Practice: File continuations from existing applications to capture claim scope now available under Desjardins that wasn't viable under prior examination standards.
Prepare for Shifted Examination Focus: With Section 101 barriers reduced, expect increased scrutiny under Sections 102, 103, and 112. Strengthen prior art analyses and enablement support accordingly.
Looking Forward
Director Squires' statement that "categorically excluding AI innovations from patent protection in the United States jeopardizes America's leadership in this critical emerging technology" signals institutional commitment to AI patent eligibility. Combined with his August 2025 memorandum establishing preponderance standards for Section 101 rejections and promised forthcoming examination guidance, the USPTO has created its most favorable environment for AI patents in over a decade.
While Federal Circuit alignment remains uncertain, Desjardins transforms the strategic calculus by shifting Section 101 uncertainty from the examination phase to enforcement. Patent applicants can now proceed with confidence that technical AI innovations won't face categorical exclusions at the USPTO, though portfolio strategies should account for potential litigation challenges until judicial doctrine evolves or legislative reform provides clarity.