Following the USPTO’s high-profile actions in Desjardins, the next question was how that guidance would play out in ordinary PTAB appeals. Ex parte Carmody provides an early answer.
The decision itself is short. The analysis is sparse. And that is precisely why it matters.
Continue Reading Ex parte Carmody: A Quiet but Important Signal on AI Eligibility Under §101
On December 4, 2025, the United States Patent and Trademark Office (USPTO) issued two memoranda addressing the use of Subject Matter Eligibility Declarations (“SMEDs”) during patent prosecution, both from Director John Squires. The first, the SMED Examiner Memo, targets the Patent Examining Corps, clarifying how examiners should treat applicant-submitted evidence under 37 CFR 1.132 in traversing subject matter eligibility (SME) rejections. The second, the Best Practices for SMEDs Memo, addresses the legal community, emphasizing strategies for maximizing the value and probative weight of SMEDs. This article summarizes both memos, offering practical insights for innovators, patent professionals, and stakeholders navigating the Section 101 landscape.
Continue Reading USPTO Issues Guidance on Subject Matter Eligibility Declarations (“SMED”)
In our earlier posts, we discussed the significance of Ex Parte Desjardins: first when the Appeals Review Panel vacated the § 101 rejection, and again when Director Squires designated the decision as precedential.
Today, the USPTO has taken the next step: it has formally issued a memo to the examining corps announcing updates to the MPEP that incorporate Desjardins directly into the Office’s subject-matter eligibility framework.
Continue Reading Further Update: USPTO Issues Memo Integrating Desjardins Into the MPEP
PatentNext Summary: Announcing the IPO’s AI Patenting Handbook V3.0, which is a newly updated third edition of IPO’s practical guide for attorneys working with AI-related inventions and technologies. It offers a clear framework for understanding modern AI (including foundation models and generative AI), drafting and prosecuting stronger AI patent applications, and navigating enforcement, global practice, and emerging AI inventorship and governance issues—designed as a day-to-day reference for in-house and outside counsel.
Continue Reading Artificial Intelligence (AI) Patenting Handbook: Version 3.0
On August 4, 2025, the USPTO issued a memo to patent examiners with the subject “Reminders on evaluating subject matter eligibility of claims under 35 U.S.C. 101.” [1]
Much has been made of these reminders and what they might signal in terms of a possible shift in how the Office treats rejections under § 101, in particular for applications involving software and artificial intelligence (AI). [2][3][4] However, it is worth considering that these reminders were not issued to the entire examining corps but instead only to three (out of eight) of the technology centers (TCs) which examine utility patents. While applicants and practitioners may be hopeful that these reminders are the beginning of a sea change in the world of § 101 rejections, especially as related to AI, they should also be concerned about the reminders’ exclusivity and what this may mean for consistency in examination standards across the USPTO.
Examiners at the USPTO are divided into TCs which are then further subdivided into art units (AUs). Broadly speaking, each TC is directed to a particular type of technology, and each AU is directed to a subcategory of that technology. For example, TC 2100 provides examination for patent applications including “Computer Architecture, Software and Information Security,” and AUs 2121-2129 and 2141-2148 provide examination for patent applications including specifically “Artificial Intelligence.” For another example, TC 3700 provides examination for patent applications including “Mechanical Engineering, Manufacturing and Products,” and AUs 3771-3775 provide examination for patent applications including “Medical & Surgical Instruments, Treatment Device, Surgery and Surgical Supplies.” [5]
The three TCs to which the August 2025 reminders were directed – TC 2100, TC 2600 (Communications), and TC 3600 (Transportation, Electronic Commerce, Construction, Agriculture, Licensing and Review) – examine most of the utility patent applications which involve AI. Consider, for example, the 163,565 utility patent applications which were filed between January 1, 2015, and December 31, 2023, which have been assigned to a TC and which include the phrase “artificial intelligence” or “machine learning” anywhere in the text of the application:
This subset was chosen because newer applications may not have been assigned to a TC yet, and older applications which recite “artificial intelligence” or “machine learning” are likely not directed to AI as it is currently understood today. Data is provided by Juristat.
As shown, 117,695 (71%) of the applications which recite AI or machine learning were examined in one of TC 2100, 2600, or 3600. But this still leaves 29% of applications unaccounted for. In particular, TC 2400 (Computer Networks, Multiplex, Cable and Cryptography/Security) received 22,586 (14%) of the applications mentioning AI which were filed during this period. And yet the August 2025 reminders were not directed to examiners in TC 2400.
This data can be further broken down. AI has become integrated into a wide variety of disciplines over the last decade; the scope of patent applications which recite AI looked very different in 2023 than it did in 2015. If we break down the data into applications filed from 2015-2019 and applications filed from 2020-2023, we see a corresponding shift in the percentage of applications examined by each TC:
These charts suggest that, as time goes on and AI becomes integrated into more fields, the percentage of AI applications examined in TC 2100 (where, as noted above, there is an AU specifically dedicated to AI) will decrease, and the percentage of applications examined in other TCs will correspondingly increase. Between 2015 and 2019, 30% of all applications which mentioned AI or machine learning were examined in TC 2100. Between 2020 and 2023, that number was only 24%. Another data point worth noting: the total number of applications mentioning AI has increased dramatically, with about twice as many being filed from 2020-2023 as from 2015-2019. If the trend continues, we might expect both more applications directed to AI and applications directed to AI being examined in a broader range of TCs and AUs.
This trend begs the question: is the USPTO equipped to handle such a rapidly evolving field? There are two important and interrelated issues at the USPTO that are relevant to this question: 1) application routing and 2) consistency in examination guidelines.
Application routing at the USPTO
Any evolution in patent application subject matter is likely to encounter some friction at the USPTO. The organization of TCs at the USPTO and the way application routing is handled are particularly relevant for newer fields like AI, which did not exist in their current form at the time that relevant procedures were put in place.
Application routing at the USPTO is determined by patent classification. Each application is classified according to a particular scheme, and then the application is routed to a relevant AU based on that classification. There are two classification schemes used for utility patents before the USPTO: the US Patent Classification (USPC) scheme and the Cooperative Patent Classification (CPC) scheme. [6] The USPC scheme has largely been replaced by the CPC scheme, and patents no longer have their USPC class printed on them, only their CPC symbol(s).
According to the USPTO website, documentation about the USPC scheme exists today only for historical purposes. [6] It is clear from the update record (called “classification orders”) that the USPC scheme has not been updated since 2013. [7] Consider what the landscape of AI looked like in 2013 compared to what it looks like today. Along with other newer fields, like quantum computing (QC), AI is not adequately captured by the USPC scheme, and because the USPC scheme is no longer updated, this will never be fixed.
One may reasonably wonder why this is a problem. The USPTO website says the USPC scheme is purely historic – why would we expect a “historic” classification scheme to adequately capture AI? Who cares?
This is a problem because the USPC scheme is still in active use at the USPTO for one key purpose: application routing. Utility patent applications are still, as of 2025, routed to the AU where they will be examined based on USPC class, not CPC symbol. A USPC class is assigned to each application specifically for this purpose. For a brief trial period in 2022, the USPTO attempted to route applications by CPC symbol. This resulted in so many misrouted applications that, rather than fix the problem, they simply reverted to the old procedure of routing by USPC class. As of 2025, there was no indication that the USPTO was planning to move away from USPC-based routing anytime soon, despite incorrect and out-of-date information still listed in MPEP § 909.
To continue to classify and route applications based on a classification scheme that has not been updated in more than a decade is, to say the least, not best practice.
Any classification scheme needs to be updated regularly in order to remain relevant, but in particular a classification scheme for patents, where the whole point is that something new is disclosed in every application, cannot effectively serve its purpose if it has not been updated in twelve years. The way that application routing is handled at the USPTO means it is difficult to determine whether the trend observed in the charts above is an intentional shift in how applicants’ and the Office are viewing and handling AI or simply a quirk of the routing system. Has the Office intended to spread AI-related applications across more TCs? Does the evolution of the field mean that applications which integrate AI into specific technologies should be considered differently than applications which are directed to AI in a purely computer science context? Or has any distinction arisen accidentally as it has become harder and harder to classify AI using a classification scheme that is more than a decade out of date?
As a specific example, should training machine learning models for use in medical or surgical procedures be routed to an AU that handles training machine learning models? Or should it be routed to an AU that handles medical and surgical procedures? Or either? Or both? The answer to this question should not hinge on how a classification scheme last updated in 2013 classifies AI. If the USPC scheme does not adequately capture the technology an application is directed to, there is no real way to assess whether the application has been routed “correctly.” This brings us to another relevant issue at the USPTO: consistency.
Consistency in examination guidelines at the USPTO
Contrary to the trends suggested by the charts above, the USPTO does not seem to take into account how AI-related applications will permeate the entire Office. The USPTO further seems to be short-sighted in the applicability of 35 USC § 101 to its entire examining corps, otherwise, why were the August 2025 guidelines issued only to TCS 2100, 2600, and 3600? These three TCs examine most of the AI-related applications that the Office receives, but not all of them. For one example, TC 2400 examines a large number, and for another example, applications which mention “artificial intelligence” or “machine learning” but also include the words “medical,” “surgery,” or “surgical,” anywhere in the text, are very likely to be examined in TC 3700:
One would expect a similar trend for TCs and AUs in other disciplines which have been quick to adopt AI methods.
Why is the USPTO issuing guidelines that should be relevant to the entire examining corps only to a subset of examiners? Examiners at the USPTO issue Office actions, not examiner actions, because there is meant to be consistency across the entire Office in how rules and regulations are applied. Providing guidance and reminders to only a subset of the examining corps does not help the Office achieve consistency in these issues, and it creates headaches for practitioners and applicants looking for clarity and consistency across Office actions. As things are, a § 101 rejection from an examiner in TC 3700, who has not received much guidance on this topic, looks different than a § 101 rejection from an examiner in TC 2600. This may make it harder to convince the examiner to withdraw a rejection because, in addition to being unfamiliar with best practices regarding writing a rejection, they may also not be familiar with distinguishing between persuasive and non-persuasive arguments for overcoming it. Conversely, this may lead to further prosecution and litigation headaches down the road if the examiner does not make a § 101 rejection when they should have due to lack of experience and guidance. Meanwhile, an examiner in TC 2100 or TC 2600 may not be familiar with the medical terms they need to understand in order to examine an application about machine learning in medical devices effectively, but as a consequence of the Office’s outdated routing procedures, they are stuck doing the best they can.
Conclusions
Many applicants and practitioners are hoping for changes in how the USPTO addresses § 101 rejections, especially as they relate to AI, and soon. Recent developments, such as the August 2025 reminders, have given those in the field renewed optimism that change is on the horizon. But the changes that practitioners and applicants see do not necessarily match what is going on behind the scenes at the USPTO. The way application routing is currently handled at the Office and the lack of consistency in guidelines issued to examiners suggests that the USPTO is not currently equipped to adapt to a field evolving as quickly as AI.
In addition to hoping for changes directed specifically to § 101, practitioners and applicants should also be looking for the USPTO to move away from USPC-based routing and to improve consistency in examination procedures if any real, intentional change is to be effected. The USPTO website and MPEP are misleading, and in some cases flat-out incorrect, about how the USPC scheme is used at the USPTO and how application routing is handled. The USPTO should provide more transparency in this regard. But at the end of the day, regardless of whether the Office is transparent with applicants about how applications are routed, it is troubling that the Office continues to rely on a classification scheme that has not been updated in more than a decade to classify patent applications. If routing by CPC symbol was so inaccurate, the reason for this needed to be investigated and corrected. The USPC scheme is not suitable for use anymore, and this will only become more of an issue as time goes on and technology evolves even further.
In the meantime, the USPTO must provide consistent guidance across the whole examining corps regarding § 101 and other statutes. Although the August 2025 guidance was provided to only three TCs, it is clear that many other TCs examine AI-related applications and deal with the associated § 101 analysis.
Perhaps a sea change in § 101 is coming, but a change in the Office’s general approach to § 101 is not enough. Real change is going to require specific modernization in practices related to application routing and examination guidelines at the USPTO. For that, applicants and practitioners may have to wait a bit longer.
References
[1] C. Kim, “Reminders on evaluating subject matter eligibility of claims under 35 U.S.C. 101,” 4 August 2025. [Online]. Available: https://www.uspto.gov/sites/default/files/documents/memo-101-20250804.pdf. [Accessed 2023 October 2025].
[2] M. Lew, “How USPTO Examiner Memo Informs Software Patent Drafting,” 8 September 2025. [Online]. Available: https://www.law360.com/articles/2383638/how-uspto-examiner-memo-informs-software-patent-drafting.
[3] R. N. Phelan, “USPTO Issues “Reminders” Supporting AI and Software Patenting and Instructing Patent Examiners on the Limits of Section 101 Patent Eligibility,” 29 September 2025. [Online]. Available: https://www.patentnext.com/2025/09/uspto-issues-reminders-supporting-ai-and-software-patenting-and-instructing-patent-examiners-on-the-limits-of-section-101-patent-eligibility/.
[4] D. Kass, “Squires Jumps Right Into Patent Eligibility Reform,” 1 October 2025. [Online]. Available: https://www.law360.com/articles/2395014/squires-jumps-right-into-patent-eligibility-reform.
[5] USPTO, “Patent Technology Centers Management,” 23 March 2023. [Online]. Available: https://www.uspto.gov/patents/contact-patents/patent-technology-centers-management. [Accessed 23 October 2025].
[6] USPTO, “Patent Classification,” 1 August 2025. [Online]. Available: https://www.uspto.gov/patents/search/classification-standards-and-development. [Accessed 23 October 2025].
[7] USPTO, “Classification Orders,” 27 March 2019. [Online]. Available: https://www.uspto.gov/patents/search/understanding-patent-classifications/classification-orders. [Accessed 23 October 2025].
When we first covered Ex parte Desjardins, we noted that the decision, which was issued just days after Director John Squires took office, could mark the beginning of a new era for AI and software patent eligibility at the USPTO. That prediction now appears well-founded. On November 4, 2025, Director Squires designated the Desjardins decision as precedential, ensuring that its reasoning now binds all patent examiners and the Patent Trial and Appeal Board (PTAB).
Continue Reading Update: Desjardins Decision Made Precedential
On October 8, 2025, the Office announced the “Automated Search Pilot Program,” a new initiative that will use an internal AI tool to conduct a prior art search before an application undergoes substantive examination. The pilot program slated to begin on October 20, 2025 offers applicants a unique, early look at an AI-generated report on potential prior art. This is the latest development in the USPTO’s ongoing efforts to integrate AI into its processes, following previous enhancements to the Patents End-to-End (PE2E) search suite, such as the AI-powered “Design Vision” tool for image searching in design patents.
For a limited number of applications, the USPTO will use its proprietary AI tool to perform a preliminary search of U.S. patents, pre-grant publications, and foreign patent documents leveraging the application’s classification, specification, claims, and abstract. The result is an Automated Search Results Notice (ASRN), which lists up to 10 of the most relevant prior art documents as ranked by the AI tool.
Similar to conducting a prior art search the ASRN provides an early glimpse into the potential hurdles an application may face, creating several possible strategic advantages. Practitioners and applicants can make informed decisions before significant prosecution costs are incurred. The ASRN is not an Office Action, and there is no requirement to respond; however, it can inform next steps:
While the potential benefits of using AI appear worthwhile, practitioners and applicants should approach this pilot with an objective lens, grounded in the current realities of the technology.
The USPTO is exploring ways to assist examiners and streamline the patent prosecution process for everyone. It is part of their broader strategy to leverage AI to improve examination.
For now, this is a limited pilot program. However, its results will undoubtedly inform the USPTO’s next steps in integrating AI more deeply into its operations. The program offers a potentially valuable, low-risk opportunity to gain an early strategic advantage in the patenting process. Practitioners with new utility applications to file in the coming months should consider whether their clients could benefit from being among the first to participate.
PatentNext Summary: The Desjardins decision, co-authored by new USPTO Director John Squires, signals a potential shift toward greater patent eligibility for AI and software innovations. By vacating a § 101 rejection and warning that “categorically excluding AI innovations from patent protection in the United States jeopardizes America’s leadership in this critical emerging technology,” the Appeals Review Panel (ARP) emphasized that eligibility should not be used as a catch-all to reject claims better addressed under §§ 102, 103, and 112. For practitioners, the decision highlights the importance of describing concrete technical improvements in the specification, tying those improvements directly to the claim language, and framing claims as technological solutions rather than abstract ideas. This marks a potentially significant recalibration of the USPTO’s approach to AI-related claims under Director Squires’ leadership.
Continue Reading Artificial Intelligence Patent Claims Get a Boost: USPTO Director Vacates §101 Rejection in Desjardins
PatentNext Summary: The USPTO issued “Reminders” for examiners in Tech Centers 2100/2600/3600 addressing §101 eligibility for software and Artificial Intelligence(AI) / Machine Learning (ML)-related inventions; while not changing the MPEP, the guidance is meant to sharpen examination practice. It clarifies Step 2A, Prong One by limiting “mental process” to what can be practically performed in the human mind—stating that AI claim limitations not performable mentally are not “mental processes”—and by distinguishing claims that merely involve a judicial exception (e.g., Example 39) from those that recite one (e.g., Example 47). For Step 2A, Prong Two, examiners must evaluate the claim as a whole to identify a practical application, giving weight to meaningful additional limitations and to improvements in computer capabilities or a technical field, even if the improvement is only implicit in the specification. The Reminders caution against oversimplified “apply it” rejections, require a preponderance of evidence for “close call” §101 rejections, and reinforce compact prosecution that fully addresses §§102/103/112 for every claim in the first action.
Continue Reading USPTO Issues “Reminders” Supporting AI and Software Patenting and instructing Patent Examiners on the Limits of Section 101 Patent Eligibility
PatentNext Summary: In a precedential decision, the U.S. Court of Appeals for the Federal Circuit reversed a district court’s §101 dismissal of patent claims relating to an automated system for dumbbell weight selection and adjustment, finding that the claims were not abstract under Alice step one and therefore are patent-eligible. The Federal Circuit held that, contrary to the district court’s conclusion, the claims included meaningful limitations which provide enough specificity and structure to satisfy § 101 even though the limitations were allegedly found in the prior art. The Federal Circuit re-emphasized the importance of considering patent claims in their entirety as a whole, which the district court improperly failed to do.
****
The U.S. Federal Circuit Court of Appeals reversed a decision of the U.S. District Court for the District of Utah, the latter of which had invalidated a set of patent claims claiming an automated system for dumbbell weight selection and adjustment, and remanded the case for further proceedings. PowerBlock Holdings, Inc. v. iFit, Inc., No. 24-1177 (Fed. Cir. 2025).
In the Utah district court, PowerBlock accused iFit of infringing U.S. Patent No. 7,578,771 (the ’771 Patent”) titled “Weight Selection and Adjustment System for Selectorized Dumbbells including Motorized Selector Positioning,” which “relates generally to exercise equipment” and more particularly “to selectorized dumbbells and to an overall, integrated system for selecting and adjusting the weight of a selectorized dumbbell or a pair of selectorized dumbbells.” ‘771 Patent at 1:15–19.
The district court held that the claims fail the two-step test and are patent ineligible because, at Alice step one, claims 1-18 and 20 of the ‘771 Patent were directed to the abstract idea of automated weight stacking and “implemented using generic components requiring performance of the same basic process”, and because, at Alice step two, claims 1-18 and 20 did “not add significantly more than the abstract idea of the end-result of an automated selectorized dumbbell.” Id. at *9.” (internal citations omitted).
PowerBlock appealed this decision to the Federal Circuit Court.
The Federal Circuit court reviewed claim 1 as a representative claim:
1. A weight selection and adjustment system for a selectorized dumbbell, which comprises:
(a) a selectorized dumbbell, which comprises:
(i) a stack of nested left weight plates and a stack of nested right weight plates;
(ii) a handle having a left end and a right end; and
(iii) a movable selector having a plurality of different adjustment positions in which the selector may be disposed, wherein the selector is configured to couple selected numbers of left weight plates to the left end of the handle and selected numbers of right weight plates to the right end of the handle with the selected numbers of coupled weight plates differing depending upon the adjustment position in which the selector is disposed, thereby allowing a user to select for use a desired exercise weight to be provided by the selectorized dumbbell; and
(b) an electric motor that is operatively connected to the selector at least whenever a weight adjustment operation takes place, wherein the electric motor when energized from a source of electric power physically moves the selector into the adjustment position corresponding to the desired exercise weight that was selected for use by the user.
Based on its Alice step one analysis, the Federal Circuit determined that the district court incorrectly concluded that claim 1 is “directed towards the general end of automated weight stacking” because claim 1 “seek[s] to claim systems comprising weight selection and adjustment systems consisting of the two or three ‘generic’ components, rather than any particular system or method of selectorized weight stacking” (Id. at *6), thereby “giv[ing] rise to a preemption problem” Id. at *7.
Specifically, the Federal Circuit court found that the district court had erroneously ignored limitations required by claim 1 when the district court did not consider the limitations reciting “an electric motor, coupled to a selector movable into different adjustment positions, and energizing the motor to physically move the selector via the coupling between the motor and the selector.” Id. at *8-9. According to the Federal Circuit, the district court was wrong to ignore such limitations “merely because [the ignored limitations] can be found in the prior art” Id. at *11.
As such, the district court did not properly consider, under Alice step one, the claims “in their entirety to ascertain whether their character as a whole is directed to excluded subject matter” Id. at *10 (internal citations omitted).
Further, the Federal Circuit warned “parties and tribunals not to conflate the separate novelty and obviousness inquiries under 35 U.S.C. § 102 and 103, respectively, with the step one inquiry under § 101.” Id. at f.3.
****
This precedential Federal Circuit decision is promising for those pursuing patents directed to mechanical automation systems. Such practitioners should attempt to draft claims that provide physical structure and interaction while avoiding functional language that could be construed as abstract. Even if some of the physical components are known, their claimed combination and interaction may still yield patent-eligible subject matter. We note, though, that pure software-based automation may face tougher scrutiny.
Additionally, when reviewing office actions or during litigation, practitioners can utilize this decision to push back on §101 rejections that ignore the claim as a whole or conflate subject matter eligibility with novelty/obviousness.
Subscribe to get updates to this post or to receive future posts from PatentNext. Start a discussion or reach out to the author, Lilian Y. Ficht , at lficht@marshallip.com or 312-423-3445. Connect with or follow Lilian on LinkedIn.