Software Patents: A Comprehensive Guide

By Alan Yomtobian, Esq. | Patent Bar #81,255

Takeaways

You can patent software in 2026. The law does not categorically exclude software from patent protection. What the law does require is that your claims be directed to a specific technical improvement rather than to an abstract idea carried out on a generic computer. If your application satisfies that requirement, it can survive examination and, in many cases, survive litigation.

The Supreme Court's decision in Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014) is the source of most of the difficulty in this area. But Alice does not prohibit software patents. It prohibits patents on abstract ideas that merely happen to be implemented by a computer. That distinction is the whole of the matter, and this article explains what it means in practice.

The Statutory Text and Its Judicial Complications

Congress wrote the governing provision in 1952. It reads:

"Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title."— 35 U.S.C. § 101

The language is broad. In Diamond v. Chakrabarty, 447 U.S. 303 (1980), the Supreme Court described it as covering "anything under the sun that is made by man." The Court has since narrowed that statement considerably, though without amending the statute.

The Court has identified three categories of subject matter that, despite the broad statutory text, are not patentable. These are called judicial exceptions: laws of nature, natural phenomena, and abstract ideas. The first two are uncontroversial. The third has generated decades of litigation without producing a workable definition.

The Supreme Court has never defined "abstract idea" comprehensively. In Alice, the Court declined to "delimit the precise contours" of the category. What we have instead is a two-step analytical test and a body of Federal Circuit case law that attempts, with mixed success, to apply it.

The Alice/Mayo Two-Step Test

The test asks two questions in sequence.

Step One asks whether the claims are directed to a judicial exception: that is, whether the claims are fundamentally about an abstract idea, a law of nature, or a natural phenomenon.

Step Two asks, if the answer to Step One is yes, whether the claims contain an inventive concept that amounts to significantly more than the abstract idea alone. Implementing an abstract idea on a generic computer, using generic computer functions like storing data, retrieving data, or displaying information, does not supply the required inventive concept.

The practical consequence of this test has been substantial. Approximately 64% of software patents challenged on Section 101 grounds since Alice have been invalidated. In 2024, the Federal Circuit decided 22 patent cases on substantive Section 101 grounds and found claims eligible in exactly one of them.

How the USPTO Tried to Bring Order to the Analysis

The USPTO cannot override Federal Circuit precedent, but it can instruct examiners on how to apply that precedent. In January 2019, Director Andrei Iancu issued the 2019 Revised Patent Subject Matter Eligibility Guidance, known as the 2019 PEG, to replace case-by-case comparison with a structured framework. The full history of that guidance and all subsequent updates is maintained by the USPTO on its Subject Matter Eligibility page.

The 2019 PEG made two significant changes. First, it defined the category of abstract ideas by enumeration rather than analogy. Under the guidance, examiners must find that a claim falls within one of three defined groupings before they may reject it as abstract: mathematical concepts, certain methods of organizing human activity, or mental processes. Claims outside those groupings are eligible without further analysis. Second, the guidance clarified that even a claim reciting an abstract idea may be eligible if it integrates that idea into a practical application, and that this integration inquiry does not require the additional elements to be unconventional.

The guidance worked initially. Section 101 rejection rates in first office actions dropped from approximately 25% in 2018 to 15% in 2020. By 2024, the rate had returned to approximately 24%.

Where the Numbers Tell the Hardest Story

The aggregate numbers obscure the unequal distribution of examination difficulty. Working Group 2120, which handles AI and simulation modeling within Technology Center 2100, saw 77% of office actions include a Section 101 rejection in 2024. Technology Center 3600, which handles business methods, e-commerce, and financial technology, is the most difficult examination environment in the patent system for software claims. Some e-commerce art units approach 100% Section 101 rejection rates, and TC 3600 generates over 76% of all Section 101 appeals to the Patent Trial and Appeal Board.

If your invention involves financial technology, insurance, or e-commerce, you are not imagining the difficulty. It is real and it is documented.

The Federal Circuit Cases That Actually Control the Analysis

Cases Finding Software Patent-Eligible

Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016) is the most important post-Alice decision for software patents. The claims were directed to a self-referential database table that stored all data entities in a single table, with column definitions provided by rows in that same table. The district court had held the claims abstract. The Federal Circuit reversed, holding that claims directed to an improvement in computer capabilities, rather than to economic or other tasks for which a computer is used in its ordinary capacity, are not directed to abstract ideas at all. The self-referential table produced faster search times, more efficient storage, and smaller memory requirements. Those were improvements to how the computer worked, not improvements to some business process that happened to use a computer.

McRO, Inc. v. Bandai Namco Games, 837 F.3d 1299 (Fed. Cir. 2016) extended this reasoning to claims involving automated lip synchronization using specific rules that evaluated subsequences of phonemes. The court found the claims eligible because the automation employed specific, unconventional technical means to achieve results impossible through the prior manual process.

Finjan, Inc. v. Blue Coat Systems, 879 F.3d 1299 (Fed. Cir. 2018) upheld claims to behavior-based virus scanning that generated a new kind of security profile data object with capabilities no prior art system possessed.

Contour IP Holding v. GoPro, 113 F.4th 1373 (Fed. Cir. 2024), deserves particular attention because it was the first Federal Circuit eligibility reversal in nearly three years. The court found claims to a dual-stream point-of-view camera system eligible as a specific means of improving the technology itself.

Cases Finding Software Not Patent-Eligible

Electric Power Group v. Alstom, 830 F.3d 1350 (Fed. Cir. 2016): collecting and analyzing power grid data is abstract.

Intellectual Ventures v. Symantec, 838 F.3d 1307 (Fed. Cir. 2016): email filtering by content is abstract.

RecogniCorp v. Nintendo, 855 F.3d 1322 (Fed. Cir. 2017): encoding and decoding facial data is abstract.

Recentive Analytics v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025), is now the controlling standard for AI and machine learning claims. The court held that claims applying established machine learning methods to a new data environment, without disclosing improvements to the machine learning models themselves, are patent ineligible. Training a known model on new data and letting it optimize iteratively is, in the court's words, incident to the very nature of machine learning. That is not an inventive concept.

The pattern is consistent: claims that recite specific technical architectures improving computer performance are eligible; claims reciting abstract data processing on generic hardware are not.

The 2025 Policy Shift Under Director Squires

In 2025, under Director John Squires, the USPTO undertook the most significant effort since the 2019 PEG to reduce Section 101 rejection barriers for AI and software patents.

The August 4, 2025 § 101 Memorandum instructed examiners in Technology Centers 2100, 2600, and 3600 that Section 101 rejections should issue only when it is more likely than not that the claim is ineligible. The memo also clarified that AI limitations that cannot practically be performed in the human mind do not qualify as mental processes. Training a neural network with billions of parameters on terabytes of data is not a mental process performed in the human mind. The National Law Review published a clear overview of the memo's practical implications in its August 2025 analysis of the USPTO guidance.

Ex parte Desjardins, designated precedential on November 4, 2025, is now binding on all USPTO examiners and the PTAB. Director Squires authored the Appeals Review Panel decision personally. The claims at issue covered a method of training a machine learning model on multiple sequential tasks while preserving knowledge from previously learned tasks, addressing the phenomenon known as catastrophic forgetting. The panel found that even though the claims recited mathematical concepts, they integrated those concepts into a practical application: the claimed technique reduced storage requirements, reduced system complexity, and preserved prior task knowledge. That was a documented technical improvement to how the ML system functioned, not merely an application of known ML to a new domain. Baker Botts published a concise summary of the decision and its designation as precedential in its November 2025 analysis.

Director Squires's opinion stated that categorically excluding AI innovations from patent protection jeopardizes America's leadership in this critical technology, and that Sections 102, 103, and 112 are the traditional and appropriate tools to limit patent protection to its proper scope.

The December 2025 SMED Guidance created a new evidentiary tool. Applicants may now submit Rule 132 declarations providing objective evidence, including performance benchmarks, comparison data, and expert testimony, that their invention constitutes a genuine technological improvement. This allows applicants to place facts before the examiner rather than relying solely on legal argument. Haynes Boone published a practical breakdown of the SMED framework and its best practices in its December 2025 client alert.

PTAB reversal rates responded to these changes. From a historical average of 8% to 12%, the reversal rate rose to 18% in October 2025 and 29% in November 2025.

Practical Strategies for Protecting Software Inventions

Lead with the Technical Improvement

The Enfish principle is the foundation of every successful software patent application: your claims must address how the computer does its job better, not the business result the computer achieves. The question is not what the system does, but what the system does differently at the level of the technology itself.

After Recentive v. Fox, AI claims require particular care. You must disclose improvements to the ML models themselves. A claim that says "we trained a neural network to optimize television schedules" is abstract. A claim that says "we modified the training pipeline to implement a specific technique that prevents catastrophic forgetting, achieving measurable improvements in the following metrics" has a defensible basis for eligibility.

Write the Specification as Evidence, Not Description

Your specification must contain three things: (1) a description of the technical problem in the prior art, stated in technical rather than commercial terms; (2) an explanation of the specific technical mechanism by which your invention addresses that problem; and (3) documented, measurable improvements compared to prior approaches, including processing speed, memory usage, accuracy, or any other quantifiable metric.

The December 2025 SMED framework makes a technically detailed specification more valuable than ever. If your specification documents that your system processed 10,000 queries in 4.2 seconds versus 11.7 seconds for the nearest prior art, you can support a Rule 132 declaration with that data. An examiner must address that evidence, not simply assert that your claims are abstract. The USPTO's own guidance on subject matter eligibility examination by date of issuance provides the full chronological record of every relevant memo and update.

Control the Art Unit Assignment

The USPTO assigns applications to art units based on terminology in the title, abstract, and claims. Applications that land in TC 3600 face dramatically worse Section 101 outcomes than those in TC 2100 or TC 2400. Lead with technical framing in every component of your application. "Machine Learning Architecture with Reduced Memory Overhead for Sequential Task Processing" will be examined differently than "System for Optimizing E-Commerce Revenue Through Machine Learning," even if the underlying technology is similar.

Reconsider the Appeal Calculus

With PTAB reversal rates doubling under Director Squires, appeals now represent a viable strategy for AI and ML claims that can leverage the precedential Ex parte Desjardins decision. The most effective appeal arguments are these: (1) the examiner characterized the claims at an impermissibly high level of abstraction, citing Enfish; (2) the examiner failed to evaluate the claims as a whole, citing Diamond v. Diehr, 450 U.S. 175 (1981); and (3) the rejection did not meet the more-likely-than-not threshold established by the August 2025 memorandum.

One procedural note: the AFCP 2.0 program expired on December 14, 2024. After-final prosecution now relies on standard after-final amendments under 37 C.F.R. § 1.116, pre-amendment examiner interviews, and requests for continued examination.

The Legislative Question: PERA

The Patent Eligibility Restoration Act of 2025, S. 1546, was reintroduced on May 1, 2025, by Senators Tillis and Coons and Representatives Kiley and Peters. If enacted, PERA would eliminate the judicially created exceptions to patent eligibility and replace them with five narrow statutory exclusions:

1. Standalone mathematical formulas not applied to produce a specified result

2. Processes that are substantially economic, financial, business, social, cultural, or artistic

3. Mental processes performed solely in the human mind

4. Unmodified human genes as found in the human body

5. Unmodified natural materials as found in nature

A critical safe harbor: any process that cannot practically be performed without a machine or manufacture would be eligible.

The bill has substantial support from the life sciences industry, individual inventors, AIPLA, IPO, and BIO. It faces opposition from large technology companies, the Electronic Frontier Foundation, and a coalition of IP law professors. The Senate Judiciary Subcommittee held a hearing on October 8, 2025, and Patently-O covered the hearing in detail. As of early 2026, the bill has not passed. Senator Tillis's announced retirement creates both urgency and uncertainty about its prospects.

The Present State of the Law, Stated Plainly

The law governing software patent eligibility is in a state of transition. The Federal Circuit's invalidation rates remain high, but the USPTO's internal examination environment has improved materially for AI and software inventors. Ex parte Desjardins, the SMED guidance, and the August 2025 memorandum represent genuine and documented improvements at the examination stage. Whether patents issued under that more favorable examination environment will survive post-grant Federal Circuit review remains to be seen.

What is certain is this: claim quality has never mattered more. The era of claiming "we do X on a computer" is over. What survives are claims that identify, with technical specificity, what the invention does differently at the level of the technology itself. Not the business result. Not the user experience. The specific technical mechanism by which the system achieves a documented, measurable improvement over the prior art.

That is the standard. Draft to it.

Frequently Asked Questions

What does a software patent cost?Software patents typically cost $10,000 to $20,000 or more for drafting and prosecution, due to the need for detailed technical specifications and claims that are resilient to Section 101 challenges. The USPTO's fee schedule provides the current filing, search, and examination fees.

How long does prosecution take?Average pendency for software-related art units is 24 to 28 months. Track One prioritized examination can reduce that to 6 to 12 months for an additional fee of $1,000 to $4,000.

Can I patent an algorithm?Not in the abstract. You can patent a specific application of an algorithm that produces a technical improvement, provided the claim is directed to the technical implementation and not to the mathematical concept itself. The Supreme Court addressed this directly in Gottschalk v. Benson, 409 U.S. 63 (1972).

Do open source contributions affect my patent rights?Potentially. Public disclosure of your invention, including open source contributions, triggers the one-year statutory bar under 35 U.S.C. § 102(b)(1). File your application before or within one year of any public disclosure.

Should I file a provisional or a nonprovisional application?A provisional application can secure an early priority date while you refine the software. But the provisional must contain sufficient technical detail to support the claims you file in the subsequent nonprovisional application. The USPTO's guidance on provisional applications explains the requirements and limitations.

More from Yomtobian Law

• How Do You Respond to a Section 101 Eligibility Rejection?

• What Is a Patent

• AI and Copyright Issues

Schedule a Consultation

Patent prosecution is fact-specific and strategy-driven. If you are developing software or AI technology and want to understand whether and how to protect it, contact Yomtobian Law P.C. for a strategic consultation.

Email: [email protected]

Phone: (516) 234-0789

Book Online: https://www.yomtobianlaw.com/book-online

Legal Disclaimer: This article is for informational purposes only and does not constitute legal advice. No attorney-client relationship is formed by reading this content. Patent law is complex and fact-specific. Consult a qualified patent attorney for advice on your particular situation.