Protecting Software & AI Innovation: What Technology Leaders Need to Know

By Stephen Durant, Of Counsel

03/18/2026

Protecting Software & AI Innovation: What Technology Leaders Need to Know

 If you’ve ever dipped a toe into the world of software or AI patents, you’ve probably heard of the Alice/Mayo test. If you haven’t, I envy you—because this test has been challenging for patent attorneys and inventors for over a decade. But good news: this guide will quickly get you up to speed, without requiring you to become a patent attorney.

 Let’s break down what’s going on with patent eligibility for software and AI in 2026—and how to survive U.S. Patent Office (USPTO) scrutiny.

The Alice/Mayo Test (AKA: “Do I Even Have a Patent-Eligible Idea?”)

The U.S. Supreme Court created a two‑step test that boils down to this:

  1. Are you trying to patent what might be considered an “abstract idea”?  Some examples are math, mental steps, and organizing human activity (including using a computer to organize human activity over the internet or with an app on a smartphone).

  2. If yes, does your invention add significantly more?
    That “significantly more” must result in a real technological improvement—showcasing an abstract idea as an innovation is not enough to merit patent-eligibility.

 The USPTO’s manual (the MPEP) tries valiantly to explain this, even including flowcharts and examples to help examiners decide whether your idea is truly a technological improvement or just using a machine for abstract brain yoga.

The Million-Dollar Question: Does Your Claimed Invention Improve Technology—or Just Use a Computer as a Prop?

This is the battleground of §101 rejections.  During real-world patent prosecution, arguments often focus on whether your patent claim includes “additional elements”—extra steps or components in the claim—that result in improving a technology and that integrate the claim’s abstract idea into a real-world practical application.  For software inventions, that means showing that your invention:

  • Improves how software works, and

  • How the improved software improves a technology

as opposed to merely using a computer as a tool to perform an abstract idea.

 Applicants and examiners may each see the same subject matter in a patent claim through a different analytical lens — one highlighting technological contribution, the other focusing on potential abstraction. An examiner interview is a common forum for resolving these differences.

New 2025–2026 USPTO Guidance: Hope for AI Innovators

Two key changes dropped recently:

A Big Court Win: Ex Parte Desjardins

 In the past, the prospects for overcoming a §101 rejection on appeal have been dismal.

 However, in Desjardins, a review panel reversed a rejection of machine‑learning (ML) claims and scolded the PTAB for calling ML algorithms unpatentable by default.  The panel in Desjardins recognized that improving how a machine‑learning model learns is a technological improvement because it improves computer functionality. That is, improved model training led to a more efficient model that enabled an artificial intelligence system to use less storage capacity, reducing system complexity.

 The case referenced Enfish v. Microsoft, in which the court explained that software innovations, defined by logical structures and processes rather than physical features, can contribute to improvements in computer technology.  This is a big deal: it means ML improvements aren’t just “abstract ideas”—they can be an improvement to a technology that can be “additional elements” in a claim that integrates patent-ineligible elements of the claim (math, mental processes, organizing human activities) into a patent-eligible practical application.

 As a result, the odds of a successful appeal of a §101 rejection in AI/ML cases have improved dramatically.  

A New Tool: SMEDs (Subject Matter Eligibility Declarations)

In addition, the USPTO now encourages applicants to submit declarations containing actual evidence that their invention results in a technological improvement. This can help build a factual record that supports patent eligibility. This combo—better case law + more flexible evidentiary procedure—opens doors for software and AI innovators navigating the eligibility maze.

 In December 2025, the USPTO issued two memos:

  • One telling examiners: “Hey, remember applicants can submit declarations with evidence.”

  • One telling applicants: “Here’s how to use that option wisely.”

 SMEDs allow you to introduce facts—not just arguments—about how your invention improves technology. This can be huge when an examiner’s view of your invention doesn’t match reality.  A SMED can be used to submit evidence, based on discussions with an examiner during an interview, to enrich your patent-eligibility arguments.

Practical Tips

Here’s your plain‑English checklist for improving your chances of §101 success:

  1. Before filing, try to predict eligibility issues.

  2. Use the USPTO’s examples to spot abstract‑idea landmines.

  3. Explain why your invention improves technology—not just why it’s new.
    This is a different inquiry than novelty or nonobviousness.  Your patent professional can put this into your patent application description section.

  4. Ask your patent professional to include claim components that serve as “additional elements.”
    They should help transform the abstract idea into a real application.

  5. For software inventions, explain how the software makes the computer better.
    Faster? More efficient? Requiring less memory?

  6. For machine learning inventions, describe how improved training improves the system.
    Your model isn’t magic (even if it appears to be)—say what’s technically improved.

  7. Highlight what’s unconventional in your invention.
    Novelty and nonobviousness ≠ eligibility—but showing unconventionality helps.

  8. Ask your patent professional to speak with the examiner before filing a written response.
    It often saves time, money and frustration.

  9. Consider using a SMED to add facts to the record.
    After an Examiner Interview that’s unsuccessful, this can make a difference -  evidence over arguments.

  10. If you have to appeal, Desjardins gives you a stronger footing for AI.
    It’s now precedent. 

 Need more assistance? Reach out and MLO can help. 

Next
Next

Your 2026 IP Readiness Checklist: What Leaders Should Do in the New Year