New USPTO Director Squires Takes an Early Position on Patent Eligibility

Just a few days into his tenure as director of the U.S. Patent and Trademark Office, John A. Squires has sent a message to the Patent Trial and Appeal Board (PTAB) to tone down its aggressive use of § 101. This message comes in the form of an Appeals Review Panel (ARP) decision,[1] written by Squires and joined by Acting Commissioner for Patents Valencia Martin Wallace and Vice Chief Administrative Patent Judge Michael W. Kim, that vacates a PTAB finding of ineligibility under § 101.

The patent application at issue was number 16/319,040 and relates to training machine learning models. During its prosecution, the examiner rejected its claims as being obvious. The PTAB heard the appeal, affirmed the examiner, but also entered a new ground of rejection under § 101.[2] The appellant requested a rehearing, which was denied by the panel.

The claimed subject matter is drawn to a machine learning model training technique that allows a model to be trained to perform two different tasks without training for the second task causing the model to become significantly less adept at the first task. The specification of the ‘040 application explains:

By training the same machine learning model on multiple tasks as described in this specification, once the model has been trained, the model can be used for each of the multiple tasks with an acceptable level of performance. As a result, systems that need to be able to achieve acceptable performance on multiple tasks can do so while using less of their storage capacity and having reduced system complexity. For example, by maintaining a single instance of a model rather than multiple different instances of a model each having different parameter values, only one set of parameters needs to be stored rather than multiple different parameter sets, reducing the amount of storage space required while maintaining acceptable performance on each task. In addition, by training the model on a new task by adjusting values of parameters of the model to optimize an objective function that depends in part on how important the parameters are to previously learned task(s), the model can effectively learn new tasks in succession whilst protecting knowledge about previous tasks.

Put another way, training a model to perform the first task causes the model’s parameters to take on certain values such that the model can achieve the goals of the first task with a reasonable level of quality. However, the model then being trained to perform a second task could result in these parameters being changed so that the model can perform the second task reasonably well but its level of quality at performing the first task is reduced. The claimed invention recites (among other features):

training the machine learning model on the second training data to adjust the first values of the plurality of parameters to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task.

Thus, the model’s ability to perform the second task can be established without having a deleterious impact on its ability to perform the first task.

The PTAB panel based its new ground of rejection on the claims allegedly reciting an abstract mathematical idea (“an approximation of a posterior distribution over possible values of the plurality of parameters”). The panel went on to state that it found “no additional element (or combination of elements) [in the claims] that may have integrated the judicial exception into a practical application.” The appellant disagreed, citing the above-quoted section of the specification and arguing that “the claimed subject matter provides technical improvements over conventional systems by addressing challenges in continual learning and model efficiency by reducing storage requirements and preserving task performance across sequential training.”

The ARP agreed. Its reasoning relied largely on the Federal Circuit’s 2016 Enfish, LLC v. Microsoft Corp. decision, which recognized that technical improvements can be found solely in software. The ARP noted that Enfish stood for the principle that claims “directed to an improvement to computer functionality versus being directed to an abstract idea” can be patent eligible. Here, the ARP determined that such an improvement was recited by the claim and supported by the specification, explaining “[w]e are persuaded that constitutes an improvement to how the machine learning model itself operates, and not, for example, the identified mathematical calculation.”

Then, the ARP goes on to make its key point:

Under a charitable view, the overbroad reasoning of the original panel below is perhaps understandable given the confusing nature of existing § 101 jurisprudence, but troubling, because this case highlights what is at stake. Categorically excluding AI innovations from patent protection in the United States jeopardizes America’s leadership in this critical emerging technology. Yet, under the panel’s reasoning, many AI innovations are potentially unpatentable-even if they are adequately described and nonobvious-because the panel essentially equated any machine learning with an unpatentable “algorithm” and the remaining additional elements as “generic computer components,” without adequate explanation. [] Examiners and panels should not evaluate claims at such a high level of generality.

However, it is with this view that the panel’s sua sponte action is most troubling, as it eschewed the clear teachings of Enfish, and instead substituted only a cursory analysis that ignored this well-settled precedent. Panels should treat such precedent with more care, especially when acting sua sponte.

At the same time, the claims at issue stand rejected under § 103. This case demonstrates that §§ 102, 103 and 112 are the traditional and appropriate tools to limit patent protection to its proper scope. These statutory provisions should be the focus of examination.

Thus, the ARP did not disturb the panel’s finding of obviousness or that the claim contained an abstract idea, but vacated the panel’s new ground of rejection because the claim integrated the abstract idea into a practical application.

Director Squires’ opinion appears to be a shot across the bow of the PTAB and the examining corps, instructing them to stop applying the § 101 analysis in a sloppy and conclusory fashion. To date, the USPTO’s application of the Alice Corp. v. CLS Bank Int’l framework has been inconsistent and problematic because it leaves almost any software claim vulnerable to being labeled ineligible, regardless of its technical contribution. This creates a system where outcomes hinge less on the actual invention than on an examiner’s or administrative judge’s subjective reading or the specific linguistic structure of the claim. The result is a test that operates more like a discretionary filter than a principled standard, denying protection to genuine technological advances simply because they are implemented in programmatic code.

But questions remain. It is unclear how significant an impact this decision will have on the PTAB or individual examiners. Also, its reasoning might not be applied to some situations, such as when the claims do not explicitly recite a technical improvement or that improvement is not explained in the specification. Further, this decision may resurrect the potency of Enfish, at least in the USPTO (Enfish had a very reasonable holding that quickly became a paper tiger).

Moreover, it is important to realize that USPTO practice and procedure has little to no influence on the behavior of district courts and the Federal Circuit. These bodies seem to have no problem repeatedly misapplying § 101 jurisprudence in the very same manner criticized by this decision.

In short, we can expect some changes at the USPTO, but not elsewhere. A small step, but in the right direction.

[1] Ex parte Desjardins, https://www.uspto.gov/sites/default/files/documents/202400567-arp-rehearing-decision-20250926.pdf.

[2] Our analysis suggests that the PTAB issues such sua sponte §101 rejections about 10% of the time. See https://www.patentdocs.org/2025/08/ptab-101-appeal-stats-for-2024-the-more-things-stay-the-same-the-worse-they-remain.html.