USPTO Director Squires’s 101 Reforms and the Growing Federal Circuit Divergence on AI Patent Eligibility

Friday, December 5, 2025 

Executive Summary

Instructor(s)
Jane Qu, Van Pelt, Yi & James
Nikola Datzov, University of North Dakota
Emily Roberts, Morrison Foerster
Harnik Shukla, Knobbe Martens

Keywords

Legal Analysis 

The USPTO’s August 2025 Eligibility Memo and Ex Parte Desjardins: A Directorial Signal That § 101 Reform Is Here

The USPTO’s August 4, 2025 guidance memorandum, addressed to Technology Centers 2100, 2600, and 3600, and the subsequently precedential Ex parte Desjardins Appeals Review Panel decision together represent the most concrete administrative action on § 101 subject matter eligibility in years — and both carry an unmistakable message from Director John Squires about the direction of examination. Shukla identified two provisions of the August memo as directly responsive to persistent prosecution frustrations: first, a clarification that the “mental process” exception must be limited to processes a person could “practically” perform in their head, excluding complex algorithmic limitations tied to specific sensors or hardware; and second, a requirement that examiners actually apply a preponderance-of-the-evidence standard when evaluating specification-based evidence submitted in response to § 101 rejections, rather than summarily ignoring it. The Desjardins ARP decision — one of only two such exercises of the sua sponte Appeals Review Panel mechanism in the USPTO’s history, Roberts explained — involved a machine learning training method application the PTAB had rejected as directed to an abstract mathematical function with no practical application; Director Squires reversed, finding that the specification’s disclosure of reduced computer storage and multi-task efficiency improvements, reflected in specific claim limitations, constituted a sufficient practical application at Step 2. Roberts characterized the decision as “a broader statement by the director that categorically excluding AI from being patentable subject matter is not going to be how things are moving forward at the patent office,” and noted his express signal that examination of AI inventions should pivot toward § 102, § 103, and § 112 rather than defaulting to § 101. Datzov framed the practical implication directly: in his assessment, “there hasn’t been a better time to try to get through patent claims that might have faced a 101 challenge, probably in the last 15 years — since Bilski at least, maybe even earlier,” suggesting that applicants who have been conservatively holding back potentially eligible claims should consider advancing them now. Datzov also noted that Director Squires’s October statement to the Senate Subcommittee on Intellectual Property — characterizing Mayo and Alice as having been “wielded as bludgeons to exclude entire classes of invention from financial technologies to artificial intelligence to diagnostics” in ways the Supreme Court neither intended nor Congress authorized — amounts to a pointed critique of the Federal Circuit’s application of § 101, even if the PTO lacks authority to override Federal Circuit precedent.

The Prosecution Strategy Dilemma: Building Patents That Survive Both the PTO and the Federal Circuit

The core strategic tension identified throughout the panel — most sharply posed by Martin’s question about what prosecutors must do to build patents that can actually be enforced at the district court level — is that the PTO’s increasingly permissive § 101 posture may produce issued patents that face invalidation the moment they are litigated. Datzov framed this as the difference between “an issued patent or a valid patent,” emphasizing that the divergence is not hypothetical: the Federal Circuit in September 2025, in the rideshare decision, explicitly stated that it “decline[s] to adopt” the USPTO’s three-prong framework derived from the January 2019 revised guidance, and that it would instead “evaluate the board’s decision under our precedent” following the two-step Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014), framework. Datzov observed that the court was not merely asserting non-binding status but was “rejecting the framework,” a distinction with real downstream consequences for claims that pass PTO examination under the agency’s guidance but would not survive the Federal Circuit’s § 101 scrutiny. Shukla offered concrete drafting advice rooted in this duality: technical improvements must be in the specification — “applications drafted 15, 20 years ago, maybe you didn’t think about that, so, having a robust specification is going to be the most important thing” — and the specification improvements must be tethered to claim language through a “technical hook” that can survive Federal Circuit specificity scrutiny. Roberts counseled a “multifaceted claim strategy” in which the claim family includes both broader independent claims that take advantage of the current PTO environment and narrower, more specifically drafted claims more closely aligned with established Federal Circuit case law, so that there are fallback positions if the broader claims are later invalidated. Qu added that the prosecution record should be built from the outset around a coherent problem-solution narrative — “you almost have to market a little bit to kind of say, ‘Well, this is the problem, but thank goodness we invented this great solution'” — so that continuations and later claim families have a coherent intrinsic record to draw upon. Wayne offered a practical examination-room observation: monitoring examiner assignment changes between office actions, reviewing the last ten to fifteen dispositions for a given examiner, and identifying patterns in what has succeeded in overcoming § 101 rejections before that specific examiner can materially affect outcomes, particularly given that junior examiners appear to be operating under more aggressive rejection mandates.

Recentive Analytics v. Fox Corp. and the Federal Circuit’s Current Line on AI Patent Eligibility

Recentive Analytics, Inc. v. Fox Corp. (Fed. Cir. 2025), the Federal Circuit’s first AI-specific § 101 decision, was widely discussed — but the panel cautioned against over-reading it as a categorical statement about AI patent eligibility. The court held, as Datzov quoted directly, “Today we hold only that patents that do no more than claim the application of generic machine learning to new data environments without disclosing improvements to the machine learning models to be applied are not eligible under 101” — a formulation Datzov characterized as limited to “a really easy case” that “is not really the case we should be paying attention to in terms of whether AI is patentable on eligibility grounds.” Shukla distilled the Federal Circuit’s operative principle from the decision: “unless you’re improving the model itself, like the black box, the application of the model is rarely going to be patent-eligible,” and observed that the court’s analysis effectively imported a prior-art-type inquiry into § 101 — something he illustrated by running the Recentive opinion through ChatGPT and receiving a well-developed § 103 obviousness analysis, suggesting these claims would have been more naturally addressed under § 103. Datzov identified two pending Federal Circuit cases — Dental Monitoring SAS v. Align Technology, No. 24-2270, and Rensselaer Polytechnic Institute v. Amazon.com, No. 24-1739 — as the decisions more likely to define where the eligibility line actually falls for AI inventions, both being fully briefed and awaiting oral argument as of the time of the panel. Qu predicted that as more AI patent cases reach the Federal Circuit, specifications will need to include substantially more detail — “what types of AI, the improvements to AI, what types of training data” — than was typical when the Recentive patents were drafted, when AI was described at the level of using machine learning to speed up processes previously done manually. Shukla also flagged a doctrinal curiosity in the current PTO guidance: at the PTO, reciting “training a neural network” at that level of abstraction does not invoke an abstract concept, but adding technical specificity such as back-propagation can paradoxically transform the claim into one reciting a mathematical concept — a dynamic that means “the more specific you get, then now you have recited a mathematical concept,” creating an inverse relationship between specificity and eligibility that has no clear analogue in Federal Circuit doctrine. Datzov predicted that historical data on AI patent allowance rates — which showed a “huge jump” following the January 2019 USPTO guidance — will bear watching as a leading indicator of whether Director Squires’s current reforms are producing measurable changes at the examination level.

Generated by AI based on the Interview/Transcript below.

Key Takeaways

• PTO and Federal Circuit now operate under different § 101 frameworks. Datzov warned that the Federal Circuit’s rideshare decision “wasn’t just saying they’re not going to be bound by the framework, they’re saying ‘We’re not gonna follow your framework,'” creating a genuine risk that patents issued under the current PTO guidance will not survive Federal Circuit scrutiny.
• Now is the best time in 15 years to advance § 101-challenged claims. Datzov predicted that the combination of the August 2025 memo and Ex parte Desjardins means “there hasn’t been a better time to try to get through patent claims that might have faced a 101 challenge, probably in the last 15 years,” and practitioners holding back conservatively drafted claims should consider advancing them.
• Technical improvements must be in the specification and anchored to claim language. Shukla advised that every specification should be audited for explicit improvements — “if they have not talked about the technical improvements, then definitely at this stage, there’s no excuse” — and that at least some “technical hook” in the claims must connect back to those improvements.
• Recentive is a narrow holding, not a broad AI eligibility bar. Datzov cautioned that the Federal Circuit’s holding covered only AI claims that “do no more than claim the application of generic machine learning to new data environments without disclosing improvements to the machine learning models,” and that Dental Monitoring v. Align Technology and Rensselaer v. Amazon will provide more consequential guidance.
• A multifaceted claim strategy is now essential. Roberts recommended drafting claim families with both broader claims calibrated to take advantage of the current PTO environment and narrower, Federal Circuit-compliant fallback claims — “a multifaceted claim strategy where maybe you have different types of independent claims that have slightly different scopes” — to hedge against the divergent standards.
• The mental process exception now has a “practicality” limit at the PTO. Shukla reported that the August 2025 memo directly addressed examiner overreach in applying the mental process exception to complex algorithmic limitations, requiring that a process be practically — not merely theoretically — performable in the human mind before being so characterized.
• Examiner-level resistance to § 101 reform will delay practical impact. Shukla observed that “for some of them, 101s become like a religion sort of thing,” predicting uneven implementation of Director Squires’s guidance across examination units, and that data suggesting reversals doubled and new § 101 ground rejections fell by over 300% in October reflects early directional movement but not yet uniform adoption.
• Interview practice is the most effective tool for guiding examiners to specification evidence. Qu identified the newly restricted interview limits — which have reduced the ability to obtain multiple interviews before an RCE — as a significant obstacle to overcoming § 101 rejections, because interviews remain “the best way for patent applicants to get over that 101 by having this kind of human conversation with the examiner.”
• The PARA Act remains a legislative wildcard. Wayne cautioned the audience that the entire current § 101 regulatory posture is vulnerable to legislative displacement — “you’re one post on Truth Social away from all of this going away. PARA Act exists, and it could be tacked on as the next rider” — making long-term prosecution strategy dependent on a politically unstable statutory baseline.

B-CLE Recording (CLE: $100) | Youtube Recording | Resource(s) | Speaker Bio(s) & Contact Info

Download the interview/transcript here!

Interview/Transcript

This interview/transcript was based on a conversation on December 5, 2025 at the 26th Annual Berkeley-Stanford Advanced Patent Law Institute. This transcript features the second prosecution panel, Navigating 101 — Understanding and navigating the PTO’s current positions, given by Nikola Datzov, University of North Dakota; Emily Roberts, Morrison Foerster; Harnik Shukla, Knobbe Martens and moderated by Jane Qu, Van Pelt, Yi & James.

[MARTIN]
So before the session start, just a housekeeping item. The discussion is great. Make sure you get the mic.

Justin’s got this side of the room. I’ll get one on this side of the room.

Just give us five seconds to get you a mic so we can keep all of this discussion for the future. cause the comments, I mean, the great thing about this room is you’re all so advanced, the comments and discussion are part of the presentation.

So keep those comments coming but just let us record them. Otherwise, you’re gonna either get to repeat them or somebody’s gonna get to paraphrase them, and that may or may not be right from there.

So appreciate that. So this panel again from last year actually the last several years, we’ve had Jane join us from Van Pelt Ye & James. Jane Qu prosecution partner and been a, been a long-term partner for the, the center here, both her and her firm.

So I’m happy to, to have her back, and thank you for leading this panel.

(applause)

[JANE QU]
So we’re gonna be discussing navigating the USPTO’s current positions on 101 subject matter eligibility. Thanks so much for joining us today. So starting as a quick introduction of all of our panelists to my left, we have Nick Datzov, who’s an associate professor at the University of North Dakota School of Law, where he teaches in the areas of intellectual property, artificial intelligence, torts, and remedies. And to his left is Dr. Emily Roberts, who’s a patent practitioner from MoFo and works in the areas of pharmaceuticals, biotechnology, chemistry, and materials.

And to her left is Harnik Shukla, who’s a partner at Kenobi Martins and has expertise in the area of medical devices. There have been some recent developments of note at the USPTO with respect to 101 subject matter eligibility, both before and, in particular, after USPTO Director John Squires was sorn– sworn in in late September of this year.

We will be covering what those updates are, their immediate effects on patent examination and at the PTAB, as well as what they might mean for patent practitioners. Lastly, we will discuss the differing interpretations of 101 eligibility at the USPTO and at the Federal Circuit and practically how these parallel standards might affect patent applicants. Um, so to start with the updates, the USPTO issued a memo titled Reminders on Evaluating Subject Matter Eligibility of Claims under 35 U.S.C.

101 on August 4th of this year. Harnik, can you tell us about the intended audience of this memo and also what the memo covered?

[HARNIK SHUKLA]
Sure. So I was actually really excited when this memo came because there’s two things in here that I wanna highlight which I’d been struggling with at the patent office. So the memo is directed to Technology Centers 2100, 2600, and 3600, of course, which we know is the business method art unit. And 2100, 2600 are more on the software side networking, communications, and a lot of AI applications are going in those art units as well. What I would have wished was if this was addressed to all the examiners because majority of my practice in, is in 3700, which is the medical device and some of the diagnostics.

And there’s, like I said, there’s two things that I thought were really interesting in the memo. The first was mental processes. Around end of last year, all of a sudden I saw every single 101 rejection was not based on math, which it used to be or a human method of organizing activities, but it was mental processes. So any time that you had some sensor data or some sort of algorithm, the examiners were just taking every single algorithmic limitation even though it was tied to specific sensors and improvements, for example, you know, improving in determination of heart rate from your optical sensors, and you have very specific algorithms.

But what was happening was every single limitation was being like, placed into a mental process. Like, a person could theoretically do this complex frequency domain processing in their head. And so the very first section of this memo says, you know, cut it out, right? Like, mental process, it has to be practical.

You have to practically be able to do it in your head. And so that’s been great. I still think it’s going to take time for examiners to understand where these practical limits are.

I have a few cases in appeal and, you know, I’m still working through it, but at least the guidance is fantastic trying to figure out, like, trying to put a limit on what is the mental process verses what’s not. So that’s one.

The other thing also that I had been struggling with was this concept of evidence. And whenever an examiner comes up with a 101 rejection, and the last several years we’ve been doing a pretty good job in writing very detailed specifications, showing the improvements you know, how something is done and talking about all the technical effects. So once, response to the 101 rejection when we would, specific citations from the specification there was nothing that came across from the examiner just summarily ignoring everything that I was citing.

And this is not attorney arguments. This is actual evidence from the specification. So there’s a section in this memo where it talks about, you know, the examiners have to follow, like, preponderance of evidence standard, like how it is for other rejections, that it’s not just, like has to be more likely than not in eligibility.

It’s not when it’s close call or all of that, So so I think that will also help because up until now, I’ve been just using other evidence for 103 and all of that in my appeals, bringing it along. But now there is something in the memo itself that talks about it. And then, yesterday I think we heard in other panels, the director came out with those declarations, right?

So it’s, the more we’re able to, you know, submit evidence and actually get it analyzed, the better the arguments will be and, you know the analysis between the applicants and the examiners.

[JANE QU]
Thank you. Going chronologically, then on September 26th of this year, Director Squires issued the Ex Parte Desjardins Appeals Review Panel, or ARP Decision which was later designated as precedential. Emily, can you tell us about what an Appeals Review Panel is and also a little bit about this case?

[EMILY ROBERTS]
Yes, happy to, Jane. Okay, so the Appeals Review Panel is a very rarely used process that consists of, or I should say comprises at least three PTAB members, and the default is the director, the commissioner of patents, and the chief PTAB judge. And this is a sua sponte convened panel by the director, and they review appeals from ex parte appeals reissue appeals, and reexamination appeals. And this isn’t something that can be, you know, requested. They don’t review or even consider requests for an ARP review, but it is something that is available, although I am only aware of two instances where it’s, you know, happened.

One of them being this recent decision ex partes, ex parte Desjardins. I’m sure I’m gonna have a different pronunciation than everyone else, I’ve heard it many different ways. But in that case, it was an application that was directed to methods of training machine learning models on various tasks.

And on appeal, the PTAB had affirmed the obviousness rejections that were raised during prosecution, but then also issued a new 101 rejection. And that was on the basis of, under Step 1, the claims being directed to an abstract idea because it was a mathematical function. And under Step 2, they didn’t identify anything in the claims, either alone or in combination, that amounted to a practical application of, you know, the mathematical formulation such that it would be patent eligible subject matter. And so after that appeal, the Director Kathi Vidal had sua sponte convened the ARP and reviewed the, the PTAB’s decision and disagreed with their characterization of Step 1 and Step 2.

Mainly, you know, the Director, I think, said it may, the claim may be directed to an abstract idea, but disagreed that it didn’t result in a practical application of the abstract idea. And in support of this the, you know, Director Squires pointed to the fact that the specification described technological improvements advantages, right, of this method. It included reduced storage in the computer, and there was also the fact that you could have multiple tasks where you’re not losing efficiency. And it remembered, you know, the first task.

And so that was some technological improvements that were not only in the specification, but importantly at least in, you know, the Director’s eyes, they were also in the claims. So there was a claim limitation that the director, you know, relied on as being tied to that technological improvement such that, you know, it was a practical implementation of the abstract idea.

And so, I guess importantly this is a much well-known decision because it not only clarified, you know, some disagreement, I guess, in what AI technologies can be patented, but it was just a broader statement by the director that categorically excluding AI from being patentable, you know, subject matter is not going to be, you know, how things are moving forward at the patent office. And that there’s going to be a change in how they think about prosecuting AI-related technologies. And they, you know, the director noted that instead of, you know, heavily relying on 101 that the proper, you know, examination focus should be on 102, 103, and 112.

And so, I expect that there’s gonna be some changes if not already being, you know, implemented at the Patent Office.

[JANE QU]
Thank you. Nick, so what do you think that the August 4th memo and the Ex parte Desjardins ARP case signal about the current direction of the USPTO stance on 101 subject matter eligibil-ability, and in particular with respect to the area of patenting AI inventions?

[NIKOLA DATZOV]
Yeah, so I think there are two examples, but I don’t think they’re gonna be the only examples we see about the direction that Emily and Harnik talked about that I think the USPTO is heading in. And I think the director so far has been pretty clear in terms of what his view is on the way patent eligibility should work at the PTO going forward, at least under his direction. Uh, you know, recently in October, in his A, ALP, AIPLA remarks, I think he mentioned that if there’s one thing he’s gonna do, it’s gonna make sure that the PTO is wide open to transformative and emerging technologies. Uh, and I think that’s signaling in everything that I’ve seen from his public remarks, as well as this guidance and including the things that we saw yesterday, uh, are, are consistent with that viewpoint, that he’s gonna try to make AI and other areas that have found some challenges in the 101 space, more open to those possibilities of, of moving into the next areas of inquiry for patentability.

You know, how far that door has been opened now, or if there is even a, going to be a door, I think we probably won’t know for a little bit, until we get to see some of the data. We’ve seen some of it so far that we can mention hopefully a little bit in our presentation later. Um, but I suspect that most examiners and, and PTAB judges are paying attention, obviously, to what’s going on And I think the sort of very public criticism in the Desjardins opinion probably is also another message that’s being sent internally to the examiners to be really cautious about what they’re doing with 101. I can’t imagine that, you know, the individuals who were part of the decision loved the sort of public reprimand in that regard, but perhaps, again, that was part of the message that the director was trying to send.

Now, I think there are gonna be some pretty significant downstream decisions as a result of the way that the PTO is going, and again, I think we’ll touch on that in our conversation today. Um, but I think in terms of, like, what the takeaways are from the public remarks, from the August 4th memo, from the Desjardins decisions is probably that there hasn’t been a, a better time to try to get through patent claims that might have faced a, a 101 challenge, you know, probably in the last 15 years, right? Since Bilski at least, maybe even earlier. And so, you know, if you’ve been holding back claims that you were worried about with 101 now might be a good time for that.

And in fact, I think again, I believe it was in his remarks at AIPLA that he said there’s gonna be a new era coming in of eligibility, at least at the PTO, right? I don’t, and I don’t think that’s gonna be true at the Federal Circuit at least I don’t have any reason to suspect it will be.

But at the PTO, there does seem to be a bit of a change in that regard. In terms of your second point about AI specifically, you know, I think that’s especially true in that space, and I don’t think it’s been an accident in terms of the two patents that he wanted to sign off on to as soon as he took office. I don’t think it’s an accident in terms of the decision he chose to publicize in Desjardins and what the subject matter there was because at the heart of all of the conversations on 101 reform that have been happening for quite a while now, but certainly in the last few years have been, you know, AI has been named as a prominent example of some of the technologies that we really need to be mindful about, and obviously is now. It’s taking greater hold in our society in a lot of different ways, and certainly I think the current administration is focused on how AI can empower certain things and legislation and that.

So I think the more that we have conversations about AI, the more it’s gonna be a focal point about what kind of innovation the PTO is promoting in that space, and I think 101 is gonna play at least at the PTO, a pretty significant role for them

[JANE QU]
Yeah.

[MARTIN]
So, because I really belong in the other room, I’m gonna bring the perspective from over there. They’re gonna say they don’t care. That room is gonna look at this and pick up their phones and start checking email, because the PTAB or the PTO is completely irrelevant to what’s gonna happen in their world. So, Okay.

Molly and I can fight about it. Yeah, I know what happens when you cite PTO regs to federal judges. They throw things at you. So, what do you need to do as a patent prosecutor?

I can get it through the PTO. How do you get?

You can make it non-reviewable afterwards, because the PTAB is essentially defined. So, yeah. So this is the discussion.

How do you then start working things through to protect them at the District Court? ’cause at the end of the day, you’re trying to get patents that can be licensed, that are enforceable, but right now, no one’s gonna be scared because of the PTO regs. So, what do prosecutors need to do going forward to build strong patents, if they’re taking these chances?

Any changes?

[JANE QU]
Yeah. Actually, you’re skipping ahead to further down my outline.

(Jane laughing)

But, I think that right now we’re seeing the trend is that, there is a, I guess, advertisement by Director Squires that the PTO is in a pro-patent-owner-friendly kind of environment, and how should we as patent practitioners, myself, Emily, and Harnik, how do we advise clients to kind of leverage this moment to try to get patents, but also with the expectations that they’re going to go through much stricter scrutiny by the time they get evaluated at the courts? Which is also a matter of statistics, because only so few cases do eventually get litigated later.

So maybe skipping to that question, Emily and Harnik, what are your perspectives and advice for clients?

[HARNIK SHUKLA]
Right. Well, I would say the first thing is getting a patent is important before your litigators get litigated. Without a patent, you’re not gonna be able to do anything.

[WAYNE]
You keep looking away, Martin.

(audience laughs)

[HARNIK SHUKLA]
I’ve been involved in many litigations, and they’re always like, “Oh, why did you do it this way?” I’m like, “Well, I got you something to do, something with a patent,” Right?

(audience laughs)

As Director Iancu said yesterday, there’s been whiplash between like, It becomes patent-friendly, then it doesn’t. But as patent prosecutors, you know what I tell clients is like, “Look, your patent’s for 20 years, “so we need to think about, how this can survive 20 years, design arounds.” There’s a whole bunch of things. But from a 101 perspective, improvements need to be in the specification.

Like when I am reading a draft from my associates, I just first do Control F and see what the word improve. If they have not talked about the technical improvements, then definitely at this stage, there’s no excuse, right? Like applications drafted 15, 20 years ago, maybe you didn’t think about that. So, having a robust specification is going to be the most important thing for claims to survive in the future because that’s where you are gonna get a lot of the evidence from the improvement.

But at the same time, you also have to think about the nexus between that improvement in the specification and some claim limitation that you can, like, tie, give that technical hook. So, keeping in mind when you’re drafting claims that do you have something which talks about how something is done, right? The Federal Circuit has repeatedly as I think Nick was talking about earlier specificity.

So, you know, you don’t need maybe the entire specificity in the claim, but some sort of hook that can tie it back to your improvement that’s in the specification. I think those things are going to be really important, like, every time you make a claim amendment, thinking about these things from a 101 perspective.

[EMILY ROBERTS]
And I wholeheartedly agree with everything you said, Harnik. My practice, it’s more in the diagnostic and nutraceutical spaces that I, you know, find myself facing 101 issues. And those are both technologies that maybe have a well-established kind of claim structure to get around 101 rejections. And that’s been established by the Federal Circuit to be patent eligible if for a diagnostic claim, you also have maybe a method of treatment.

And that’s giving you some improvement. And so of course best practice is to have in your application, you know, disclosure related to the improvements preferably with data kind of showing measurable concrete advantages.

And that can be pointed to during prosecution, backup positions that can be included in your claim, you know, if need be that are tied to those advantages even more specifically than maybe initially in your claim set. But what I think I’m interested in seeing in view of these recent changes is, is how patent prosecution practice with respect to, to those, you know, non-AI related technologies kind of evolves. You know, will people be more risky in pursuing diagnostic claims that may have not been you know, passing muster with the Federal Circuit? I think to hedge your risk, you’re probably gonna have to have a multifaceted claim strategy where maybe you have different types of independent claims that have slightly different scopes based by virtue of different claim language and kind of see what gets through the Patent Office.

And then hopefully you have some claims that are more closely aligned with established case law in you know, the Federal Circuit District Courts that would be fallback positions in the event that something. You know, is not favorably viewed from the PTO, which I guess the guidance is not binding, so they kind of disregard it. But at least you’d have other claims that you could fall back on

[NIKOLA DATZOV]
If I might jump in on that point. So I think Wayne’s question’s a really, really important question, and I think if you’re a patent practitioner, specifically a prosecutor right now, I think you really need to be thinking about talking with your client on what their strategy is in getting a patent. Are they trying to get an issued patent or a valid patent? And I’m not trying to coy here because I, I do think there is a difference between those two things sometimes, at least based on right now the guidance that’s been issued, right?

And I agree with Wayne’s point. For those of you that aren’t perhaps convinced or familiar with it, right, there has been a divergence between the USPTO’s guidance on how eligibility should be applied, at least since the January 2019 guidance, and what the Federal Circuit law has been. In some cases, they align, but in some cases, they don’t. And so we’ve seen these repeated statements from the Federal Circuit that say, as Emily mentioned, “The guidance isn’t binding,” but we saw an even sort of more, I think, pointed statement just recently in September, in the rideshare decision from the federal circuit, which said, well, let me make sure I get the quote right here, but I don’t have it handy, but the board, the board used the USPTO’s 2019 revised patent subject matter eligibility guidance and the October 2019 update, which uses a three-prong framework.

We decline to adopt this framework, is not binding on this court and instead evaluate the board’s decision under our precedent, follows a two-step test set out in Alice. Now, maybe I’m reading too much into that sentence, but, but the Federal Circuit here wasn’t just saying they’re not going to be bound by the framework, they’re saying, “We’re not gonna follow your framework,” which is another maybe way of saying they’re rejecting the framework. And so, I think, we’ll see what clarity comes out of that.

That was a decision that just came out in September. But I think for a while, the Federal Circuit judges have been signaling, “We’ve got some issues with the way that you’re approaching eligibility at the PTO. We look at it differently.” And so to answer Wayne’s question about like if you’re, if you are thinking about past the stage of getting it issued, which of course is a threshold question.

If you don’t get it issued, patent you don’t have to worry about the Federal Circuit. But if you do and your clients are thinking about enforcing it or this is a particularly valuable patent, then my answer to Wayne’s question about what you should be doing is, is reading the Federal Circuit cases, right? Because I think if you’re passing muster under the Federal Circuit’s test, you’re almost certainly gonna be passing muster under the USPTO guidance as well.

I’m not sure the reverse is true, and I think that’s kind of the divergence that we’ve seen a little bit in some of the cases come out. So again, you know, it’s nice to have a more favorable, I think wide open sort of door at the PTO in terms of getting your clients the patents that they need.

And when I was in practice, I talked with clients. I didn’t prosecute, but I was a litigator and gave advice as well on some of the 101 issues specifically. And when I talked with clients, they were like, “I just need a patent. I don’t really care about all the 101 issues you’re talking to me about that might materialize 10 years from now.”

And I get that, right? But I think strategically, if you’re thinking ahead and you’re trying to, to plan for those and avoid some of those issues, you’ve gotta look at the 101 case law from the Federal Circuit and, and make sure that you’re being compliant with that, irrespective of what the USPTO guidance is saying right now.

[JANE QU]
Yeah, and I think the–

[NIKOLA DATZOV]
Question, I don’t know if we’re taking questions.

[WAYNE]
Yeah, I just wanted to add that I think it’s interesting that Squires is definitely thinking about this problem. You mentioned his AIPLA remarks where he talks about the three-legged stool of eligibility, and he says, “We already have the tools we need.” So he, he functionally says like, “I don’t care about the other Federal Circuit decisions.

I love NFEISH.” And it’s really interesting remarks that comes up in Desjardins as well that he’s clearly trying to set up a framework where he can have his cake and eat it too. But it’s obviously something that’s on his mind that I’m sure we’re gonna see more statements that are gonna give advice to patent holders to, on how to address that problem.

[NIKOLA DATZOV]
I agree with you, and I think he’s been pretty open about that. You know, another remark that he shared which kind of stood out to me when I was going back through and preparing for today again, in October, this was not in the AIPLA remarks. This was in an October 10th statement to the US Senate Subcommittee on IP on the Committee on Judiciary, and he said there, “Supreme Court’s later decisions in Mayo and Alice have been widely misinterpreted.

Properly read, they do not narrow eligibility beyond these” And I’m quoting here, “They do not narrow eligibility beyond the established judicial exceptions, laws of nature, natural phenomena, abstract ideas, yet in practice these decisions have been wielded as bludgeons to exclude entire classes of invention from financial technologies to artificial intelligence to diagnostics. That is not what the Court,” capitalized, meaning Supreme Court, “Intended, and it is certainly not what Congress ever authorized.” And it’s not entirely clear to me if he’s talking about the way the USPTO has applied these standards, but I think probably a fair reading is that he’s criticizing the Federal Circuit’s approach and how they’ve interpreted the law.

And of course, I think the problem we probably all realize is that irrespective of his views on what the law should be, an administrative agency as the PTO doesn’t have interpretive authority to decide what the law should be. It has to follow the Federal Circuit’s case law on it.

Now, you know, again, I think at a high level they align. I don’t think Director Iancu is saying, “Let’s get rid of the two-step framework.”

But how it’s being applied to the USPTO, I think that’s probably gonna draw some questions about that. And again, I think this is where we’re gonna start to see some interesting data come out, right? So some of the data that I’ve seen early on, Professor Dennis Crouch had a really great post on Patently-O in October that you can take a look at if you missed, and he shared some data about the number of 101 rejection, new ground rejections by the PTAB as well as the number of 101 reversals in October of 2019 compared to prior months, and essentially the reversals doubled and I think the new ground rejections were reduced by like more than 300%. So I think already we’re starting to see some of the impact of Director Iancu’s view on 101.

And again, I think that’s gonna create a little bit of a different path that, that Wayne was pointing out too in terms of, well, what are the district courts gonna think about those issued patents when they get there?

[HARNIK SHUKLA]
I think there’s also a aspect of resistance within the examiner community and some judges as well. I don’t think everybody’s on board with Director Iancu’s vision.

You know, don’t have that much data yet, know,I did a rehearing request based on Desjardins and You know, that didn’t go anywhere, it was similar arguments, and the examiners are still, you know, waiting to figure out like, you know, how much are they rolling back the 101 rejection. So I think, you know, for some of them, 101s become like a religion sort of thing, so it’s gonna be hard for them to not think about it, and I think it will take some time before we see practical effects even at the Patent Office.

[JANE QU]
Emily, has that been true of your experience too since Dieter Danze? I don’t know if you’ve applied it in your practice.

[EMILY ROBERTS]
I have not. I’m sure that my colleagues who work more closely in the intersection of AI and life sciences have maybe seen some changes in at least the examination of their pending applications. I have not.

I think there are just larger policy changes being made at the PTO right now, and I have seen, you know, other effects in my practice because of those, but not specifically with respect to 101.

[NIKOLA DATZOV]
And on the AI space, I think we do have a little bit of precedent about this. Again, we don’t have a ton of data from what’s happened with the guidance after Director Squires took office, but we did see a little bit of this, so I wrote an article a couple years ago about AI patents and eligibility, and part of what I wrote about in there was comparing the data of allowance rates with what the AI, what the PTO had characterized as AI patents in its AI dataset. And if you looked at the allowance rates, there was a really interesting sort of jump in when you compare, like, what the allowance rate was in 2014 through 2018, and there was a decline in the allowance rates of what the USPTO categorized as AI patents. Now, I think there’s some questions about what falls in that bucket to begin with.

But taking the bucket as it is, at least as it was framed by the PTO, there was a decline in the allowance rates of AI patents until 2018, and then in 2019, which again, remember what happened in 2019, we received the guidance from the USPTO, there was a huge jump comparatively speaking, in the allowance rates of those AI patents. And so, I think it’ll be really interesting to see, once we get some more data about examination records, what’s gonna happen with AI allowance rates going forward.

[JANE QU]
Yeah, I mean, AI patents, is such a broad category. I do wonder if there’ll be eventual splits, breakdowns.

You know, there’s AI in terms of using a third-party AI frontier model in application, which is, I think, a lot of the basis of the technology that we’re seeing, or, you know, are we massaging the data that goes into the models? Are we post-processing?

Are we validating? Are we trying to curb hallucinations? So I think as more cases are being filed, it’ll be interesting to eventually get kind of a breakdown of, you know, what we call AI inventions, you know, which slice of these different AI types of inventions are maybe getting more traction through the Patent Office and which are facing more uphill battles.

Um, and I think, sorry, to go back to the disconnect between the, you know, Federal Circuit’s view of or stricter view on 101 in regards to patents, and the, I guess, more f– patent-owner-friendly view of 101 on patents at the PTO, I would say that, you know, the patent’s value once issued, is not zeroed out by the stricter scrutiny that we might expect from the courts. And it’s sort of a discount to the overall value, so it does kind of make the value of the patent a little shakier, knowing if it’s like in a more embattled area, and what we might expect if it ever gets litigated.

But, you know, certainly having an issued patent, like Harnik says, there’s a lot of value in that that isn’t zeroed out just because it might face s– some trouble if it gets litigated, which is also a big if. And I think it also speaks to the value of having continuations open you know, being able to, you know, have the initial, at the point of filing, having the story of, you know, we face these technical problems and this is the technical solution we came to overcome those problems, really kinda sell your invention and your application. You know, you almost have to market a little bit to kind of say, “Well, this is the problem, but thank goodness we invented this great solution, made everything so efficient and so much more accurate,” et cetera, because you really need to kinda go back to that well later as you’re prosecuting, you know, hopefully more continuations in your family to have more options open.

And also maybe other potential options to consider for clients is, or for applicants, is the idea of, you know, I think typically, you know, we talk with inventors, it’s a more specific idea and then our job is to help broaden that idea a bit. But I think there’s also value in not only getting the broader independent claims if possible, which is to also get the independent claims that are a bit more specific and do actually cover exactly what the invention is.

And sometimes, maybe the original product, and sometimes that maybe better aligns with the original problem-solution sort of story that is also in the specification. So that way, you kinda have a few cases maybe in the family that would be harder to defeat altogether in litigation for 101 issues. Yeah, I guess I don’t know, Nick, I know we kinda touched on this already, but what do you think are the biggest differences between how the PTO has characterized, you know, examining 101 and how the Federal Circuit has interpreted that.

[NIKOLA DATZOV]
Well, I think one is the framework, right? So the way that the eligibility analysis proceeds at the PTO is just different from the Federal Circuit, and I think that’s part of what the Federal Circuit has pushed back on, including in the rideshare decision most recently. But it’s hard to say without using specific examples, right?

Where the divergence really kinda gets more concrete. I think specificity is another point that really has been emphasized more at the Federal Circuit than perhaps has been in the guidance, and I don’t know if we’re gonna see that under Director Squires going forward. But I think that’s a really important part that I always encourage folks who are drafting patents to think about, is not only can you demonstrate an improvement to the technology that’s at issue, but also how broadly have you drafted your patent claims? Because preemption has been a pretty I mean, was a key concern at the Supreme Court when it was deciding the 101 cases, I think we’ve seen it mentioned at times at the Federal Circuit as well.

And so, I think the Federal Circuit really is thinking about how broad are your claims, so even though you might have something that’s really novel, you might have something that does in fact, at least generally speaking, improve the technology. But if you have framed your actual claim scope in a really broad way, without actually, as Harnik talked about, incorporating the specifics from the specification about the way the technology works, at least in some way, into the claim, right? Again, that doesn’t mean I think that you have to include all the language and in the claim language itself.

I had when we were litigating a 101 motion one time the patentee literally argued to the judge and said, “Well, look how many words are on the page. Certainly this is gonna pass 101.”

And the judge we were in Delaware, kind of laughed at that, and I think rightfully so. It’s not about the number of words that you’re putting in there.

It’s about the specificity of the claim and how broad it is, and how that’s tied to the improvement that you’re saying you’ve made in the technology itself. And so again, I mean, I think then you perhaps walk yourself into some claim construction issues if you leave out a lot of words. But you can, I think, incorporate some of the things that you put in the specification into the claim language in order to highlight that you’re claiming something that is something more specific, something more narrow than just a broad abstract idea about how to functionally do something which is gonna catch the attention of the District Court and Federal Circuit, for sure.

[JANE QU]
So my last question maybe goes out to the whole panel and it’s in regards to the Federal Circuit’s Recentive Analytics versus Fox Corp case that came out this year which taught us that AI inventions are patentable, but simply using AI to do something more efficiently is itself not going to confer eligibility. So what do you think are the takeaways, you know, we have from that case with respect to patenting AI inventions going forward?

Uh, maybe we start with Harnik.

[HARNIK SHUKLA]
Sure. And, you know, and just for fun last night, I took that decision, put it in ChatGPT, and asked it, “How would you analyze this under 103 instead of 101?” I think everyone should try it. It was fascinating. It came up with a very good analysis of how that claim could be invalid under 103, and because when you read the decision, a lot of it is about, like, “Well, you know, this was known, people were doing this,” and all of that. And then at the end, like, “Well, and that’s why it’s abstract.”

It’s a very tough analysis when you’re trying to import, like, you know, prior art-type Analysis into Section 101. So I think that those type of cases would have been better analyzed under for-hard rejections than 101, but it is what it is, and I think the, well, the way I read that case, and Nick and I were arguing about that a little bit earlier, I think Federal Circuit is saying that unless you’re improving the model itself, like the black box the application of the model is rarely going to be patent-eligible. Like, you know, if you’re just receiving some data, training it and then applying it to some output. So I think that’s going to be a much tougher bar that the Federal Circuit has as compared to the Patent Office, has a very interesting, also doing something very interesting where they’re saying that training a neural network, you recite it at that level of specificity, doesn’t recite an abstract concept.

But now if you start adding details like, you know, it includes back-propagation or then all of a sudden you’re adding math to, so the more specific you get, then now you have recited a mathematical concept. So I think the Patent Office is definitely going to be broader I think the claims that were in Recentive would probably pass now at the Patent Office under the new guidance, but definitely doesn’t at Federal Circuit.

[NIKOLA DATZOV]
Yeah. I mean, we were chatting about Recentive a little bit, and I think Harnik and I have maybe a little bit of a different view in terms of how narrowly the Federal Circuit was looking to foreclose that. To me, I think Recentive was a really easy case, actually, at the Federal Circuit.

It’s not really the case we should be paying attention to in terms of whether AI is patentable on eligibility grounds. There’s a couple of other cases that, I gave a presentation on 101 and AI stuff just a couple months ago at a different practitioner conference so there’s a couple that I’ll put on your radar, happy to talk with others. But they’re fully briefed at the Federal Circuit. At least as of last I checked, they’re still waiting for an oral argument date.

But Dental Monitoring SAS versus Align Technology 24, 22, 70, and Rensselaer Polytechnic Institute versus Amazon.com which is Docket Number 24-1739, I think those are going to be more interesting decisions about where the line is on AI than Recentive was. I think Recentive was, again, like a fairly straightforward case, and if you read the conclusion from what the judges said there, they said, “Today we hold only that patents that do no more than claim the application of generic machine learning to new data environments without disclosing improvements to the machine learning models to be applied are not eligible under 101.” I don’t think, that means that if you have, you’re using AI in a novel way, a specific application that does more than just abstract idea, don’t think the Federal Circuit was saying you can’t patent that. But I think it’s saying that if you’re not improving the machine learning algorithm itself, you’re gonna draw some scrutiny, and you better be able to explain what it is that you’re doing more so than just, “Oh, well, we’ve done these things before.”

We’re just now doing them with AI.” That’s not gonna pass muster. I think that’s going to be really problematic, and I think that’s essentially what Recentive was.

[JANE QU]
Yeah. I think the patent at the heart of the Recentive case was also relatively short in the specification. I think that at the time it was written, AI was described to the point of we used to come up with these network maps manually which was very complicated, but now we can use machine learning to speed up that process, which would make it better. And I think that in 2025 or even in recent years you probably add a lot more detail when you’re describing AI in the specification. So, while it’s the first I think, AI-related case on 101 to come out of the Federal Circuit, it certainly won’t be the last, and I think we’ll start seeing more cases with much more detail around what types of AI, the improvements to AI, what types of training data.

I think they will give us a better picture of what it reaches, I guess, the line of eligibility for the Patent Office, I mean, sorry, for the Federal Circuit in terms of 101 eligibility as more cases get litigated. So, Yeah, I’d say that this is maybe, you know, as it’s always true, it’s not always maybe the best patents that kind of become the case law but sometimes you read a specification and the results seem, you know, a little bit understandable. So, I guess we’ll have to see as more cases reach the Federal Circuit. Yeah, do folks have any questions?

Yes, Ernie?

[ERNIE]
One of the things that seems open with Squires is identifying the technology being improved as a route for showing that something’s patentable under 101. How, what luck have you had and what approaches have worked best with examiners in getting them to actually look that deeply into the specification? The rule of thumb seems to be that there’s not enough time allowed for an examiner to read your specification while they’re examining your application, given, you know, the pressures of, for productivity. So, I find it quite difficult after they’ve done a 101 rejection to get them engaged in what the technology being improved is.

And a second question, have you heard any rumors that they’re using an AI to write 101 rejections?

(audience laughter)

[JANE QU]
It’s funny about the you know, we were just mentioning how the PTO under Iancu is now so much more pro-patent and one new rule that kind of cuts against that picture is the more strict limitations on the number of interviews that the examiners can grant per, you know, non-final and final office round. And before, you can kind of get away with potentially requesting an interview at both non-final and final, and potentially to get two interviews in before you have to file an RCE pay more money basically. And having those interviews is the best way to kind of you know, guide the examiner to the exact, pinpoint them to the exact portions of the specification that, you know, you think they ought to look at to you know, understand your invention. And especially when you’re trying to point them to a point in the specification that touts some sort of benefit, technical solution which is understandable that they’re not going to have enough time to read, like, an 80-page document to find and to locate the three paragraphs that best describe that.

And so, yeah, that is kind of unfortunate now that we’re losing the ability to have more interviews, which I think is the best way for patent applicants to, you know, get over that 101 by having this kind of human conversation with the examiner. But, I think maybe interviews, you know, when you can get them is further the best way to kind of walk the examiner through, you know, saves them time as well. Yeah. We’ve gotten an invitation recently from Tarek, the director in 36 under his charge of interviews, to work on reviving interview practice, so cross your fingers that, that one-hour limitation will go away in the next year.

For sure. Yes.

[WAYNE]
Yeah. So you know, on this point of interviewing, there’s, there’s two things that I’ve started to do more recently. Look at the examiner on the office action, because I noticed in about a third of my cases in three different technology centers, in about a third of cases, the examiner has changed between the first Office Action and the second office action. It may not be that you’re getting the same philosophy behind the rejection, and a lot of the newer examiners are required, as you know, to give a rejection on every single– It seems like they’re required to be more strict in rejecting, so there are a lot of junior examiners, which means you need to get the SPE or the supervisor in the interview.

Interview. So, that’s one thing I noticed.

The second thing is go into one of these tools and look at the last 10 or 15 dispositions of an examiner when you get an office action, and look at what is being successful with that specific examiner to overcome 101. I notice some examiners have, like, magic words or something, where you have the word in the specification, you cite it in the claim, somehow they withdraw the rejection.

You don’t have to put anything about the case law or the rules, and if you’re comfortable that that’s compliant with the Federal Circuit cases, you’re okay. But I’ve noticed this, using this data has been helpful, because examiners are changing and their tendencies are changing.

[HOST]
That’s a great point.

[NIKOLA DATZOV]
Can I offer one quick remark? And again, like, I think to Wayne’s point, like, if you haven’t looked at the case law recently, I really encourage you to do some of that, and I think that can help in your persuasion with the examiners, too. I know that that’s not usually where prosecution attorneys are probably living is in litigation space, but I think that’s gonna inform the viewpoints that your clients are thinking about. Bit of a shameless plug.

I wrote an article with Jason Rantzen this year that examined every single Federal Circuit decision that we’ve had since March of 2012 through December ’23, and we opened up our data set. That’s posted publicly.

The link is in the article. And so, if you’re like, well, how do I even find Federal Circuit decisions? We’ve linked all of them for you. They’re categorized by judges and by different things, and so there’s ways that you can find that information now, I think, more readily.

But I hope that everyone’s taking that into account, because I think it’s gonna help with your clients’ long-term strategies.

[WAYNE]
Wonderful, and just to follow up on what Director Iancu said yesterday, you’re one post on Truth Social away from all of this going away. PARA Act exists, and it could be tacked on as the next rider. So, with that, thank you all.

(audience applauding)