USPTO Review Panel Backs Off on AI Patent Eligibility in Recent Appeal
A recent Patent Trial and Appeal Board (PTAB) decision, Ex parte Desjardins (Appeal 2024-000567), highlights an important shift in how the United States Patent and Trademark Office (USPTO) is approaching artificial intelligence (AI) inventions—where an “AI invention” can be characterized as an improvement to how the machine itself operates, and not, for example, a mathematical calculation. This case involved an application directed to methods of training a machine learning model to learn new tasks without “forgetting” old ones—a technical solution that reduced storage requirements and improved model performance.
Initially, the Patent Trial and Appeal Board (PTAB) affirmed the examiner’s obviousness rejections and went further by introducing a new rejection under Section 101, saying the claims were nothing more than abstract math on a generic computer. That reasoning, if left standing, could have put many AI-related inventions at risk. The applicants sought rehearing, and in a rare move, the USPTO Director convened an Appeals Review Panel (ARP) to revisit the decision. The Patent ARP made clear that AI and software innovations remain patent-eligible so long as the application clearly identifies technical improvements to existing technologies.
The ARP vacated the Board’s new Section 101 rejection, finding that the invention did not simply claim “math on a computer,” but rather described a meaningful improvement to how machine learning models operate. While the claims remained rejected over prior art, the Panel’s decision signals that the USPTO will not automatically dismiss AI and software innovations as ineligible. Instead, the focus is shifting back to the traditional tests of patentability: novelty, non-obviousness, and enablement.
For businesses and innovators working in AI, this is encouraging news. It means there is a clearer path to protecting advancements in training models, optimizing performance, and reducing system complexity—so long as applications are drafted thoughtfully. The key takeaway is that successful AI patent applications should emphasize the technical improvements achieved by the invention. Describing how your approach improves efficiency, reduces costs, or enhances system capabilities will go a long way in overcoming eligibility challenges.
Applicants should also be mindful that the PTAB sometimes raises new eligibility issues even if they weren’t part of the original examination. Anticipating those arguments early and addressing them in the specification can save time and expense on appeal. And while applicants cannot directly request ARP review, strong arguments on the record may make it more likely that the Director will intervene when necessary.
In short, the Desjardins decision offers reassurance that AI innovations remain protectable under U.S. patent law, provided that the application makes a clear case for how the technology improves the way computers actually work. For companies investing in AI, this underscores the importance of strategic patent drafting that highlights real technical benefits.
Contact the authors of this Client Alert or your Butzel attorney to learn more, including if you would like to discuss how this decision might affect your portfolio or upcoming filings.
Daniel Vivarelli
202.454.2841
vivarelli@butzel.com
Mitchell Zajac
313.225.7059
zajac@butzel.com
Priya Dalal
202.454.2830
dalal@butzel.com
