For years, businesses developing software and artificial intelligence (AI) have struggled with the murky waters of U.S. patent eligibility. The challenge has often centered around Section 101 of the U.S. Patent Code, which was designed to exclude abstract ideas, laws of nature, and natural phenomena. Unfortunately, courts and examiners have frequently applied it in ways that blocked legitimate innovation in the software industry.
Now, in what could mark a historic shift for software patent applicants and AI developers, the U.S. Patent and Trademark Office (USPTO) has issued new internal guidance. A memorandum released on August 4, 2025 provides direction for examiners on how to apply Section 101, particularly in evaluating claims in software-related fields.
The big takeaway? Examiners are now discouraged from rejecting AI and software patent applications solely under Section 101, paving the way for stronger and fairer intellectual property protections.
Why Section 101 Has Been So Controversial
Section 101 has long been a source of confusion and frustration when applied to software. Starting with the U.S. Supreme Court’s 2014 Alice decision, and continuing with over a decade of subsequent application and re-interpretation, the USPTO’s Section 101 examination practice has been inconsistent and unpredictable. The incredible advancements in AI technology now bring patent applications involving AI models, neural networks, and machine learning into this area of deep uncertainty.
Over the past year, AI related patent applications have seen increased rejections under Section 101 as “abstract” even when they delivered tangible technical solutions to real-world challenges. This uncertainty has discouraged many innovators from pursuing patent protection, and have left others uncertain as to the viability of their efforts to protect their valuable innovations.
The 2025 Memo: A Shift Toward Practical Application
The new USPTO memo emphasizes reasonable, evidence-based analysis. Examiners are reminded to:
- Avoid labeling complex AI-driven tasks as “mental processes” unless a human could realistically perform them with pen and paper.
- Distinguish between claims that truly recite an abstract idea versus those that simply involve one.
- Consider whether the invention contributes a practical application or technological improvement, especially in how it improves computing or another technical field.
- Refrain from rejection unless there is more than 50% certainty of ineligibility.
Optimistically, this memo marks a cultural shift in favor of software implementations of AI technology. Instead of defaulting to rejection, examiners are now encouraged to take a holistic look at what the invention accomplishes.
Why This Memo Matters for Businesses
For software companies and AI developers, the memo is more than a procedural update—it’s a strategic opportunity to secure stronger patents.
The USPTO now recognizes that AI-driven processes such as training neural networks or real-time data analysis are not “mental processes”. Applicants can also expect greater consistency in how claims are evaluated under the Alice/Mayo framework.
Even more importantly, if your invention improves the performance, efficiency, or security of computing systems, it now stands a stronger chance of patent approval. The memo even tilts the benefit of the doubt toward applicants, ensuring that uncertain cases are less likely to be rejected outright.
Practical Tips for Applicants Seeking Software Patents and AI
While the memo is encouraging, businesses must still be strategic in how they draft and present their patent claims. To strengthen your application:
1. Showcase technical depth. Move beyond what your software does and clearly explain how it works, what problem it solves, and why it improves the system or field in which it operates. This includes describing the AI models used in the invention, as well as how those models are trained.
2. Avoid generic computer language. Claims that simply say “do this on a computer” without technical depth are more likely to be rejected. Show the architecture, algorithms, or data structures involved, including detailed descriptions of how AI models are constructed and arrive at their outputs.
3. Leverage examples. The USPTO has released several AI-related subject matter eligibility examples that illustrate how to frame your invention in a way that’s likely to succeed.
4. Emphasize practical applications. Make sure it’s clear that your invention does something meaningful in the real world, not just conceptually or theoretically.
The Bottom Line: A More Innovation-Friendly USPTO
For years, the tech community has pushed back against restrictive interpretations of Section 101. The USPTO’s 2025 memorandum signals that those concerns are finally being heard.
While the memo does not rewrite the law itself, it reshapes how Section 101 is applied, creating a more innovation-friendly environment for AI and software patents.
For businesses, this is a critical moment. If you’re developing AI or software-based products, now is the time to revisit your patent strategy. Stronger, more predictable protection could provide the competitive edge your innovation needs to thrive.