USPTO Internal Memo Suggests that Patenting AI Inventions May Become Easier

The USPTO has issued an internal memorandum that may make it easier to patent software, in particular AI-related software inventions.

In recent years, the USPTO has found certain software inventions to be patent-ineligible abstract ideas. In the new internal memorandum, the USPTO has directed patent examiners to exercise greater care before rejecting AI-related software technologies for this reason. While the memo is styled as a “reminder” of existing policy and law, its emphases and examples send a clear indication that the Office wants to dial back aggressive eligibility rejections, particularly in AI and machine learning domains.

The memo, dated August 4, 2025, came from the desk of USPTO Deputy Commissioner for Patents Charles Kim, and is titled “Reminders on evaluating subject matter eligibility of claims under 35 U.S.C. 101.” Distributed to examiners in Technology Centers 2100, 2600, and 3600—those art units covering software, computer architecture, communications, and business methods technologies—the memo offers applicant-friendly subject matter eligibility guidance that signals a welcome reprieve to applicants amidst a USPTO eligibility climate that has tightened considerably in recent years.

Not all AI patent claims are a “Mental Process”

The memo states that “The mental process grouping is not without limits,” and that “Examiners are reminded not to expand this grouping in a manner that encompasses claim limitations that cannot practically be performed in the human mind.” Specifically, with respect to AI-related inventions, the memo states that “Claim limitations that encompass AI in a way that cannot be practically performed in the human mind do not fall within this grouping.”

It should be noted that these statements are taken nearly verbatim from the USPTO’s 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence. Still, the Office’s reemphasis of these points in this new reminder memo should be seen as a strong statement in favor of applicant-friendly limitations on unchecked malleability of the “mental process” abstract-idea grouping.

Not all AI patent claims “recite” an abstract idea

The memo reemphasizes the distinction between claims that actually “recite” an abstract idea, and may therefore be patent ineligible, from claims that merely “involve” an abstract idea, which do not trigger further eligibility analysis and should be deemed patent eligible.

Specifically, the memo points to the contrast between its Eligibility Example 39 versus its Eligibility Example 47.

The new memorandum attempts to address the tension between these two examples by distinguishing between generally reciting “training a neural network” (as in eligible Example 39) versus specifically invoking mathematical concepts used for training, such as “backpropagation” or “gradient descent” (as in ineligible claim 2 of Example 47).

Practice Insight:

The argument that an AI-focused patent claim does not recite an abstract idea (in other words, the “Prong One” pathway to eligibility) has historically been received frostily by examiners. The more viable approach has generally been an argument that recited abstract ideas are practically applied (the “Prong Two” pathway). So, the reemphasis in this memo of Prong One may revitalize a previously difficult pathway to patent eligibility for AI inventions.

And, from a strategic standpoint, a revitalized Prong One pathway could open the door for broader AI claim coverage. This is because positioning a software claim for eligibility under Prong Two can often require reciting narrower technical details in the claim, whereas a claim’s positioning for eligibility under Prong One can benefit from broader framing (as shown by Example 39).

AI claims must be analyzed “as a whole” for eligibility

The memo reemphasizes that claims should be analyzed “as a whole” for eligibility, stating that “limitations should not be evaluated in a vacuum, completely separate from the recited judicial exception,” and that “analysis should take into consideration all the claim limitations and how these limitations interact and impact each other” when evaluating whether the exception is integrated into a practical application.

Practice Insight:

The requirement to analyze claims as a whole has sometimes been sidestepped by examiners who state that a claim is being analyzed as a whole, but who, practically speaking, divide claims into atomic words and phrases, deeming half of the isolated words and phrases to be “abstract ideas” and deeming all remaining isolated words and phrases to be “recited at a high level of generality.” The memo clearly attempts to push back against this kind of analysis.

AI claims are eligible if they reflect an improvement to technology

The memo reiterates that AI claims may be deemed eligible if they reflect an improvement to computer technology or to another area of technology.

First, the memo reminds examiners to “consult the specification to determine whether the disclosed invention improves technology or a technical field.” The memo further states that the specification “does not need to explicitly set forth the improvement,” but must at minimum “describe the invention such that the improvement would be apparent to one of ordinary skill in the art.”

Second, the memo cautions examiners “not to oversimplify claim limitations” to improperly characterize a claim as merely invoking the idea of applying an abstract idea.

Here, the memo contrasts the use of software technology to improve a patent-ineligible business process against the use of software to improve “technology or a technical field.” The memo at this point provides an interesting footnote regarding the Federal Circuit’s recent decision in Recentive Analytics, Inc. v. Fox Corp. (see our full summary of that decision here), contrasting the patent‑ineligible Recentive claims against the patent-eligible claim 3 in USPTO Example 47.

Practice Insight:

Historically, examiners have sometimes given credence to the specification’s explanation of the technical advantages provided by the invention. If this memo encourages more examiners to follow this practice, then applicants can preemptively strengthen their position against eligibility rejections by clearly explaining in the specification how their technology represents a concrete improvement to existing technology.

Tie goes to the applicant

Finally, the memo reminds examiners that a rejection should only be made “when it is more likely than not (i.e., more than 50%)” that the claim is ineligible, and that rejections should not be made “simply because an examiner is uncertain as to the claim’s eligibility.”

While this final point of emphasis is not substantively groundbreaking, its inclusion in the memo emphasizes the applicant-friendly tone of the memo as a whole, signaling that Office leadership would like examiners to exercise caution and discretion in issuing eligibility rejections in “close call” cases, particularly in the AI technology space.

Takeaways and Practice Tips

This memo is a signal that eligibility rejections, particularly for AI technology, should be expected to decrease under the new USPTO administration. While adoption by examiners will inevitably vary, and implementation of changes will take time, this memo signals an applicant-friendly shift in a critical technology area.

Applicants preparing new applications in the AI technology space should ensure, in light of this new memo, that specifications explain, robustly and at length, the technical advantages of as many different aspects of the invention as possible. This was already best practice, but if the specification’s explanation of technical advantage is going to become more important during the examination process, then emphasis here should be redoubled.

And, for applicants facing current rejections for lack of subject matter eligibility, particularly in the AI space, delaying and extending response timelines should be considered to whatever extent practicable, in order to allow time for implementation and potential further adjustments to take shape at the patent office.