[powerpresss]
My co-host Ken Suzan and I are welcoming you to episode 175 of our podcast IP Fridays! Today’s interview guest is Bruce Dearling, patent attorney and partner at Hepworth Browne in the UK, and we talk about how non-technical features must be considered when assessing inventive step of patents at least according to recent decisions of the UK supreme court and the Unified Patent Court.
UK Supreme Court Emotional Perception AI Limited
But before we jump into this interesting interview, I have news for you:
On May 20, 2026, the Swiss Federal Council adopted the fully revised Patent Ordinance, which will enter into force on January 1, 2027, together with the revised Patent Act. In the future, the Swiss Federal Institute of Intellectual Property will prepare a mandatory search report for each application; applicants can choose between a partially examined version and a full examination that assesses novelty and inventive step. The full examination costs an additional 300 Swiss francs, and renewal fees will increase by a total of eight percent over the 20-year term.
On May 19, 2026, Asus entered into a licensing agreement with the Wi-Fi multimode patent pool managed by Sisvel, thereby ending all ongoing infringement proceedings. Sisvel bundles standard-essential patents in the pool from, among others, Atlantia, ETRI, and Mitsubishi Electric.
On May 18, 2026, the UPC Local Chamber in Düsseldorf rejected Align Technology’s application for a preliminary injunction against its Chinese competitor Angelalign. Angelalign may continue to sell its clear aligners within the UPC jurisdiction. Our partners Dirk Schulz, Ulrich Storz, and Wanze Zhang, together with Arnold Ruess, successfully represented Angelalign.
The U.S. Patent and Trademark Office (USPTO) announced midweek that, since October of last year, it has invalidated or is seeking to invalidate approximately 10,500 trademark applications and registrations in eleven administrative orders. Reasons include forged attorney signatures and the fabrication of non-existent filing requirements. This stems from ongoing abuse of the U.S. trademark system, primarily by non-U.S. applicants, which can lead to conflicts with validly registered trademarks for legitimate businesses.
On May 12, 2026, the British Court of Appeal overturned a lower court decision that would have required Nokia to grant interim licenses for video coding patents. The court found that Nokia’s license offer to the Taiwanese manufacturers Acer and Asus had already been made on RAND terms.
In May, the U.S. Department of Justice (DOJ) filed a brief in the ongoing Corteva v. Inari litigation, expressing antitrust concerns regarding certain patent practices in the field of plant breeding. This marks the first time the agency has actively intervened in a biopharmaceutical patent dispute with implications for seed innovations.
Episode 175 of the IP Fridays podcast was a conversation I will not forget quickly. My guest Bruce Dearling, partner at Hepworth Brown in the UK and a patent attorney for 36 years, took a case through every level of the British court system up to the Supreme Court and, in doing so, fundamentally changed patent law for AI inventions in the UK. The case is called Emotional Perception, and its effects reach well beyond British borders.
Below I summarize the key points from our conversation. The full episode is available at IP Fridays.
A. What Is the Emotional Perception Case About?
The underlying invention concerns artificial neural networks. Specifically, it relates to a method of closing what is called the semantic gap at the output of a neural network. That sounds abstract, but the idea is straightforward: a neural network always produces an output that does not fully correspond to what a human would actually expect or feel. Closing that gap brings the system closer to human perception and human expectations.
Bruce Dearling drafted this application himself and filed it at the UK Intellectual Property Office (UKIPO). The Office rejected it as excluded subject matter, characterizing it as essentially a computer program as such. The legal basis for that rejection was the Aerotel decision from 2006.
The case then went to the High Court, which found in favor of the applicant. The Court of Appeal reversed that decision. Then the UK Supreme Court stepped in and changed everything.
B. The Aerotel Test and Its Flaws
Since 2006, the Aerotel test had been the standard British method for assessing whether an invention falls within the excluded categories under patent law. It was a four-step approach: construe the claim, identify the actual contribution the invention makes to human knowledge, ask whether that contribution falls solely within excluded subject matter, and finally check whether the contribution is technical in nature.
The problem Dearling described in our conversation is that Aerotel reverses the logical order of the analysis. You start with the contribution and only then ask about the exclusions under Article 52 EPC. The UK Supreme Court described Aerotel in its judgment as “unsound law” and overturned it. The EPO’s Technical Boards of Appeal had previously called Aerotel “disingenuous,” which at the time led to a public dispute between the British courts and the Boards. With the Emotional Perception ruling, that conflict has now been resolved in favor of harmonization with the EPO.
C. What the UK Supreme Court Decided
The Supreme Court made two central findings. First, the exclusion of computer programs “as such” is overcome as soon as a claim includes any piece of hardware. It does not matter whether that is a processor, a memory module, or any other component. The threshold is deliberately low. Dearling described this as the “any hardware” approach, which aligns fully with the EPO’s position following G1/19.
Second, and in Dearling’s assessment the more important finding: when assessing inventive step, the invention must be considered as a whole. The Court introduced what it called an “intermediate step,” an analytical stage in which the interactions between all features of a claim are examined before the question of inventive step is addressed. Non-technical features cannot simply be struck out if they contribute to the overall technical effect of the invention.
D. Inventive Step: The Intermediate Step
This is the heart of the judgment. In EPO practice, Dearling said, it happens regularly that examiners strike through features they consider non-technical and thereby fail to assess the invention’s inventive step correctly. A recent Technical Board of Appeal decision, T 1249/22, already criticized this approach: a claim directed at a technical solution to a problem can be patentable even if the underlying problem is non-technical in nature.
Dearling recalled a remark made by a Board of Appeal member at a hearing he attended years ago: “We understand that examining divisions can operate with a degree of mental laziness and that it’s too easy to throw too many things out of the basket when considering the issues of inventive step.” That quote stayed with him because it names a structural problem that the intermediate step now addresses directly.
The British method for assessing inventive step is the Pozzoli test, which differs from the EPO’s problem-solution approach. The Supreme Court explicitly retained Pozzoli because the problem-solution approach, in its view, is structurally infected with hindsight reasoning: you already know the invention, you work backwards to formulate an objective technical problem, and then you ask whether it would have been obvious for the skilled person to arrive at precisely that solution. Dearling sees this as a source of unfairness toward genuine inventions.
E. Alignment with the Unified Patent Court
In April 2025, the Court of Appeal of the Unified Patent Court issued a decision in Abbott v. Sinocare (APP_000000901/2025, judgment of 17 April 2025). Dearling pointed out that this decision uses language and reasoning strikingly similar to the UK Supreme Court’s Emotional Perception ruling of February 2025.
That is significant because the UPC is bound neither by UK courts nor by the EPO. The overlap suggests voluntary convergence. Dearling reported a conversation with a person close to the EPO, whom he did not name, who used the word “permissive” to describe the UK Supreme Court’s approach and indicated that the EPO might move toward it. Whether and how quickly that happens remains to be seen. What is clear is that the UPC, as the new European patent court, is setting its own standards, and the question of how to handle non-technical features in inventive step assessment is now being asked at multiple levels simultaneously.
F. Implications for the EPO and Practice
The EPO is not directly bound by the ruling. It is an administrative body, not a court. Dearling is nonetheless optimistic that change is coming. On one hand, external pressure is building: when the UK Supreme Court and the UPC articulate similar principles, convergence becomes hard to resist. On the other hand, Article 27.1 TRIPS requires all contracting states to make patents available in all fields of technology. Examiners routinely striking non-technical features from AI claims and rejecting them on that basis sits uncomfortably with that obligation.
For the underlying application in the Emotional Perception case, the ruling has a pointed consequence. The Supreme Court did not grant the patent itself; it referred the matter back to the UKIPO for reconsideration under the intermediate step. The Office’s subsequent response was, in Dearling’s words, unconvincing. He suspects the Office is attempting to reintroduce the Aerotel test through the back door. As a last resort, he has not excluded a judicial review, a procedure that does not simply challenge the substantive decision but holds the Comptroller General of Patents to account for whether the Office is deliberately circumventing the Supreme Court’s direction on the intermediate step. That is, as Dearling put it, “a nuclear option,” but one he would not rule out if the evidence in the file already suggests the Office is in contempt of court.
There is also an international dimension. Singapore’s Intellectual Property Office launched a public consultation shortly after the ruling, asking whether Singapore should adopt the Emotional Perception approach into national law. That is British soft power operating in real time within the Commonwealth.
G. Three Takeaways for Patent Practitioners
At the end of our conversation I asked Bruce Dearling to distill the most important practical points. His first takeaway: make sure the claim contains hardware. This applies not only to UK and European applications but is simply good drafting hygiene. Without hardware in the claim, the application remains exposed.
The second takeaway concerns the description. Anyone filing an AI invention needs to explain clearly which function is achieved by which piece of hardware, circuit, or software. Not as boilerplate, but as a complete technical account that describes the real-world effects. Dearling’s experience is that practitioners who write the claim first and fill in the description afterward run into trouble.
The third takeaway emerged from the conversation itself: how the EPO assesses inventive step for AI inventions is not a settled question. It is worth following the development of UPC case law and any shifts in EPO practice closely. Anyone advising on AI patent applications today needs to know these arguments.
H. Conclusion
The UK Supreme Court’s Emotional Perception ruling is not a British footnote. It has declared the Aerotel test dead, introduced the intermediate step that brings non-technical features back into the inventive step analysis, and set off a convergence movement that is already visible at the UPC and still pending at the EPO. For everyone working in AI patent practice, whether in prosecution, examination, or counseling, this ruling is required reading.
Rolf Claessen: Our interview guest on IP Fridays podcast is Bruce Dearling. He has been in the IP field and a patent attorney for 36 years and is partner at Hepworth Brown in the UK. Thank you very much for being on the podcast.
Bruce Dearling: My pleasure, Rolf. Thank you for inviting me.
Rolf Claessen: All right. We just met at the INTA annual meeting in London. And you talked about the UK Supreme Court case where you were involved. And the core questions were whether non-technical features would be considered when assessing inventive step of patents. Can you briefly summarize this case?
Bruce Dearling: It’s a bit more than that. It started — I actually wrote the case. And I prosecuted it through the patent office. The patent office rejected the case for being excluded subject matter. So pretty much the excluded subject matter provisions in the UK are nearly identical. They’re as near as practical to the language of the EPC, so those of the European Patent Office — Article 52.2. But again, they apply as such.
The actual technology relates to artificial neural networks. And the invention related to a very clever way of what is termed closing the semantic gap at the output of the neural network. So that means that in a neural network, there is always a discrepancy between the output of the neural network in terms of what it’s telling you you should be thinking essentially, and what reality is. So if you can close the semantic gap, then you align the neural network or the artificial intelligence system to better reflect human knowledge or human reactions and human expectations. So that’s really what the invention is about. There’s no point in going into too much detail with it — that’s the way it is. It’s very clever.
So the UKIPO rejected this because they said it was essentially a computer program excluded from patentability as such. And they used a decision which is called Aerotel, which has been around since 2006. And that decision has caused considerable consternation and tension between the EPO Technical Boards of Appeal and the UK courts. Aerotel was described as being essentially disingenuous by the EPO Technical Board of Appeal. And the UK courts pushed back and said, you don’t know what you’re talking about. So that’s where it fell apart.
So that’s where they rejected it for essentially being a computer program as such, possibly with a bit of business methods thrown in as well. But let’s leave that for the time being. So the case then went to the High Court and at the High Court, we won. The judge said, actually, it’s not a computer program. Neural networks aren’t computers. They’re not programs themselves. There’s more to them than that. And the invention as claimed is not excluded from patentability as such.
The UKIPO obviously weren’t very happy about that because they liked their Aerotel case and so they appealed it. And they appealed it on several grounds, including a new one, which was that it was a mathematical method. The Court of Appeal decided that the UKIPO was right and that we were wrong, so we lost the case.
So we then went to the Supreme Court. Well, actually, they denied us an ability to go to the Supreme Court. The court said no appeal. We went — actually, no, I think there is a bigger issue here — because we realized, or I realized at that point, that the work that we were doing was much broader than this. It requires real consideration of what an invention is at a fundamental level. So not only exclusions, but how inventive step is applied. And these issues were built into the case from the very beginning.
And they sort of — I wouldn’t say crept up on the court as we went through — but they became more and more prominent to the extent that ultimately, when we made an application to the Supreme Court, the Supreme Court went, yeah, we’ve got some issues here. We want to hear the full arguments on why this is not excluded from patentability, why Aerotel is potentially bad and how we more or less try to align ourselves with the European Patent Office. So that’s essentially what happened. And the Supreme Court hearing was last July. It took them the thick end of eight months to come out with a decision, which was issued in early February, at which point the entire legal landscape in the UK changed because they said we were right. The Patent Office doesn’t know what they’re talking about.
Aerotel is bad. It’s unsound. That’s what they described it as — unsound law. It needs to be removed and we’re going to harmonize with the European Patent Office.
So before I — I’m just going on a bit of a rant here, standing on my soapbox telling you what you already know. But the Aerotel test essentially was — it was a four-step test, past tense. So you firstly had to construe the claim. That’s pretty straightforward. Then you actually had to identify the actual contribution. This is what they said — identify the contribution. Really in this aspect, you’re asking what, as a matter of substance rather than form, the inventor has added to human knowledge. So that’s what they said the contribution was. And then they said, the next step in Aerotel was to ask, well, does that contribution fall solely within the excluded subject matter field or realm? And then they said, well, if you get through that question, then you check the actual contribution or the alleged contribution to see whether it’s technical in nature.
So that’s the Aerotel test as it was. And what the Supreme Court in their unanimous final decision said was that Aerotel at best jumbles up the order. It reverses the logical order of the analysis by starting with the contributions and then addressing the Article 52 exclusions. And then finally it goes back to what the technical nature of the invention is about. So they really went, no, we don’t like any of this stuff. It’s bad, it’s stupid, it puts the cart before the horse.
So, in the intervening period between finding the case and actually seeing it progress all the way to the Supreme Court, we obviously had the G1/19 decision from the EPO Enlarged Board. And they basically said that they are going to validate any hardware as the approach. And that’s essentially what the UK also went with. The UK Supreme Court said we’re going to say that the threshold of patentability — or the exclusion to patentability — is simply overcome by the inclusion in a claim of any piece of hardware, whether it’s a processor or a piece of memory or whatever. It doesn’t matter. Any hardware makes the invention a technical invention. So it’s a really low threshold to consider.
And they then went, well, actually, if we now align and harmonize with the European Patent Office sensibly, then we need to look at how we assess inventive step, which is the other thing that we raised with the Supreme Court. In fact, we probably raised it at other times and in all the other instances as well, but it came to a head at the Supreme Court. So the Supreme Court then also went a bit further and said, well, actually, whilst we do like the global approach to assessing inventive step for all fields of technology — whether it’s chemistry or biotech or electronics or software or AI — we use a test called Pozzoli. So that isn’t problem-solution. We don’t like problem-solution. We think it’s not codified in the European Patent Office. It’s just a mechanism that the EPO has come up with to try to objectively assess inventive step. We don’t particularly think that’s appropriate. We like our approach called Pozzoli.
That’s it. So we’re going to say with Pozzoli, however, in order to actually understand — particularly in the context of mixed inventions having technical and non-technical features — it’s necessary for the examiner to undertake the so-called intermediate step, where you have to look at the interactions between features within a claim. The invention is defined by the claim. That’s what the act says. That’s what everyone understands. It’s the invention defined by the claim. So you look at the claim features and then you have to understand the interactions that take place. And even if they are between technical and non-technical features, if they bring about an overall technical effect when you consider the invention as a whole, then your claim should be good and you can assess it for classical inventive step.
So that’s really where we’re at. There’s a lot to unpack there already. It’s probably a podcast in its own right, but that’s the positive history of where we’re at. And I can keep going if you wish me to for a second and talk about why I think this is — we’ll just contrast it quickly with the problem-solution approach at the EPO and COMVIK. So for inventions in the computer-implemented field, they use COMVIK and the problem-solution approach.
The Supreme Court said, as I said, they don’t like problem-solution. I think the problem-solution issue is that it is also inherently pre-baked with hindsight because you have to look at the invention and then step back and exclude those features which are common. And then you formulate a problem based on the function that the claim achieves. And then you’re asking whether or not it would be obvious for a skilled person to arrive at the claimed invention, having been given that hindsight-developed problem. So COMVIK is not great by any means.
And we know from a practical perspective that examiners are only too willing to look at a claim and simply line through features which they believe are non-technical, whereas they don’t actually look at the interaction of those features in the context of the claim as a whole.
There is also a decision — very recent one actually, about a year ago — T 1249/22, where the Technical Board of Appeal told the examiners and the examining division, you cannot do this. It’s okay to have a claim directed towards an invention in a non-technical field, as long as the invention is directed to a technical solution of that problem. I think it’s paragraphs 11 and 12 or 10 of that decision that are worth looking at. But they’re saying that in all fields of technology, it doesn’t matter as long as the technical solution is about technology — therefore, you should be able to obtain a patent as long as there is a realistic and appropriate technical effect. Be careful actually, Bruce — I don’t mean technical contribution, I mean technical effect. There’s a reason for that distinction.
Rolf Claessen: The non-technical features are nevertheless used to assess inventive step in the UK now after this decision, right?
Bruce Dearling: Yes, that is the intermediate step. The decision says you must look at the invention as a whole. It’s the important thing. There are a couple of issues that arise out of this. The first one is that you have to provide context for the invention.
The Supreme Court never provided any specific guidance about how we deal with the intermediate step or what the exact test is, which is in some respects fine. It seems to be fairly clear that you just have to engage your gray matter — your neurons — to work out what is going on in the real world. And once you work out what’s going on in the real world, what the benefits are, then you look at whether or not the actual implementation of the invention fundamentally has a technical flavor to it, which is not just coding, not just simple coding, but it does something smarter. There’s a real technical impetus. There’s a technical effect.
Now that actually brings me onto something I’ve postulated or said. I think the intermediate step will follow something like what I’ve termed the holistic character test, which essentially is: work out what’s going on in the real world. Then once you’ve worked out what’s actually being achieved, what the benefits are, what the invention’s concerned with, then you ask the question, how am I achieving it technically? And how is there a technical effect? How does the technical effect arise?
That brings out a couple of issues. The first one is that it’s actually about the word “contribution” because it depends on how the word is used. So if you look at head note one in COMVIK, it uses the word “contribute” — how the non-technical feature contributes to the invention. So that’s an additive inclusive concept. The UK IPO historically, and arguably at the moment today whilst they’re trying to retrain their 400 examiners — which this has caused them to have to do — their idea of contribution is this backward-looking concept.
So technical contribution and technical effect, I think — although we mix them up and interchange them — are distinct. Technical contribution: you’re looking backwards. Technical effect is what you look at when you look forward into what’s going on. So this is subtle — it’s really subtle, but it’s important. And once you realize that you are actually looking for the technical effects, then you’re on much safer ground. It’s much more objective in terms of the assessment. This might be somewhat contentious, because it’s the way I’m looking at this, but I’ve been working on this a long, long time and thinking about it for probably decades, worryingly so. So technical contribution and technical effects are probably not the same, where they are interchangeably used to mean the same thing within existing decisions.
Rolf Claessen: And in the beginning you said, now that Aerotel is dead basically, it’s more harmonized with the EPO’s approach. But what I take from the discussion now is that maybe — especially in view of the problem-solution approach — it’s not fully harmonized with the EPO’s approach at the moment, right? Or did the UK Supreme Court get something wrong, or was that a desired outcome from your point of view that this is not so completely harmonized with the EPO?
Bruce Dearling: Well, the EPO — the any-hardware solution is fully harmonized, no doubt. So it’s now a question of inventive step under Article 56 or Section 3 of the Act. The EPC nowhere mandates the use of problem-solution. And we know that there are many different ways of actually assessing inventive step, including the concrete elaboration test from last year and problem-of-invention approaches. So there are numerous ways of assessing inventive step. So the UK says, “Pozzoli — we like Pozzoli.” Interestingly, I had a discussion with someone I probably can’t mention. They’re saying that the UK approach may actually be more permissive now. It might even influence how the EPO operates. So they may move away from COMVIK towards more of a Pozzoli approach, which basically says this:
You identify the notion of the skilled person — step one. You identify the common general knowledge of that skilled person — step one B. You identify the inventive concept of the claim in question, where you construe it if you can’t work out what it is. You then identify what the differences are. And then you ask the question, is it obvious to the skilled person, given knowledge of the common general knowledge?
This is entirely not artificial because, as I said beforehand, when you look at problem-solution, you are formulating a problem by backtracking from what the claimed invention is to a situation where you say, well, these are the common features and I’m going to project a problem to try and solve. Now that is already tainted with hindsight reasoning. It’s not safe, it’s not thoroughly objective. There is an inherent problem with this which sees good inventions cast by the wayside. Although it’s a preferred mechanism, it’s not fully baked. There are situations where examiners are inherently lazy, or they just simply use something like the requirements specification argument, which is just factual. It just demonstrates that they can’t be bothered to actually argue it properly or think about what the invention is. Sorry to any examiners listening to this, but this is just my personal view, that sometimes there are problems.
I’m reminded of a quote from an EPI hearing I was at a long time ago, where the Legal Board of Appeal member said: “We understand that examining divisions can operate with a degree of mental laziness and that it’s too easy to throw too many things out of the basket when considering the issues of inventive step.” Now that one has stayed with me because you think — did someone just say that? And the answer is yes, they did. But it just goes to show that there is some tension between the TBA and the examining divisions, and they don’t always get it right.
Rolf Claessen: So there might be a small difference now between the UKIPO’s future approach of assessing inventive step and the EPO?
Bruce Dearling: Yeah, it might do. But the other interesting thing here — and thank you for pointing this out, I hadn’t entirely caught up with it, I’ve been traveling beforehand and I missed some of the UPC case law. So the UPC case law — in, was it — yeah, we talked about that.
Rolf Claessen: Yeah. There was a decision in April, Abbott versus Sinocare.
Bruce Dearling: Yeah, 901 of 2025. So a Court of Appeal decision from the UPC. It was APP_000000901, I believe, 2025. Decision 17th of April, hearing 27th of March.
The UPC is not bound by — it’s a court. The European Patent Office is not a court, it’s an agency that administers and looks after the administrative rule of law. So the fact that this decision came out from the UK Supreme Court in February, and you see almost identical language used in the UPC decision, suggests that there is some alignment here, or some convergence in thought. Now, whilst the UPC decision also references G1/19 and uses problem-solution, there is enough — you’ve got to bear in mind that high-level courts do look at each other’s decisions. And this is really a question of influence and the desire to converge. So the fact that they’ve done this at this time is quite interesting.
Again, I can’t quote someone directly from the EPO, although I would love to. They were saying — at a very high level — and they used the words “converge UPC practice towards UK Supreme Court practice on interpretation of the law.” So this may actually be happening in real time. Again, it would be wrong to actually refer to anyone by name, but it’s an observation that when I looked at the case, I can see why this is going ahead. And I can see why the judiciaries — they want to maintain independent judicial controls. They won’t reference the UK Supreme Court decision, not least because we’re not in the UPC. But if you look at the arguments in sections 106 and 107 of the UK Supreme Court’s Emotional Perception decision and head note one, you go — wow, this is very close.
Rolf Claessen: Very close and nearly identical wording. Yeah. And the UPC also now uses non-technical features for assessing inventive step. Is that a problem for the EPO that has historically been aggressive in throwing out non-technical features for inventive step analysis?
Bruce Dearling: Well, I think they really need to get to the situation — I don’t know — this holistic character test that I’m sort of proposing, where you really have to think about what the invention is achieving, and then look at how it’s technically being achieved. And then if you look at that again in the context of that other decision I mentioned — T 1249/22 — it says something like, in the case of an invention that amounts to a technical implementation of a non-technical method, provided the non-technical method does not contribute to the technical character of the invention. The board validated the approach of identifying the non-technical method and then goes through and says it’s patentable. There are decisions like this which suggest that examining divisions have to give it a bit more thought, because the Technical Board will realize that to satisfy the WTO requirements — which pretty much everyone is bound by — Article 27.1 TRIPS, which requires that you protect all fields of technology. And that means whether it’s data processing or business methods, because business methods can be patentable so long as they are implemented on a technical basis. That essentially seems to be what T 1249/22 is saying, although it doesn’t explicitly say “allowing business methods.” The exclusion is only “as such.” So does this decision, in combination with the Supreme Court case and the movement of the UPC, say: well, actually, let’s look at this properly? It requires objective assessments, not just superficial “let’s strike through that feature because I don’t like it, it looks non-technical.”
Rolf Claessen: So are you hopeful that the EPO is adjusting and will reshape their case law in view of the UPC decision and the UK Supreme Court decision?
Bruce Dearling: It’s a bit unfortunate that the corresponding UK case at the EPO was dropped by the applicants, because it was heading towards an examination hearing at the examining division. It would have gone to the TBA, and I’m sure it would then have gone from the TBA to the Enlarged Board. I’m pretty sure that’s the case. There is another case from the same client which will probably argue the same thing because the specs are almost identical. It’s just lagged in time. So is it going to change? I hope so, because I think the EPO have got it wrong — more often than not in this field. Well, maybe not more often than not — they get it wrong more times than they should do.
Would I like to see it changed? Yes, I would, because I want the examiners to actually think about the technology as opposed to just — oh, it’s not — I don’t want to engage the gray matter. That serves no one. That doesn’t serve technology. That doesn’t serve industry. These patent rights are there for a reason. They are property rights. I’m referring to the award of the 2025 Nobel Prize for Economics — they are a core driver for society’s development. So the 2025 Nobel Prize was for something called creative destruction — the replacement of old technology with new — and it’s based on the patent paradigm. So all this stuff is coming to a head now. It’s just a question of how quickly the EPO actually catch up, and maybe they have something to catch up on.
It’s just understanding that the examiners have to start to think. As I said, we’ve got the issues at the UKIPO where they’re going to have to retrain 400 examiners.
Rolf Claessen: Yeah, right.
Bruce Dearling: The Emotional Perception case wasn’t granted by the Supreme Court. They referred it back to the patent office for consideration under the intermediate step. So the patent office produced a response that I would describe as — I’d say arguably — not well reasoned, which I’ve filed the response to, which basically says you don’t really know what you’re talking about. What really worries me a bit is that I think they’re trying to introduce the Aerotel case through the back door. It’s backsliding. It’s a mechanism for trying to apply it in a different way or a different context, which would be wrong.
I think they believe that the applicant will appeal this if they get a bad decision — they will appeal it back to the courts again via the High Court, Court of Appeal, Supreme Court route. I say maybe not. I say maybe the client will file what they call a judicial review, which is a nuclear option. That’s when you actually hold the Comptroller General of Patents to account and get full discovery of whether or not there’s internal documentation showing that they are deliberately circumventing the direction of the Supreme Court on the intermediate step. This is basically holding them to account and saying: if you’re not applying the intermediate step appropriately, you are in contempt of the law. So judicial review is a really serious thing to do, but it’s certainly something I would not exclude from consideration.
We’ll see what happens. It’s not saying we’re just going to go through the courts and make them decide on this. We’re going to say you’re wrong. And there’s already enough evidence in the files to suggest that they are probably in contempt of court and they’re not applying the intermediate step appropriately. They may not know any better at the moment — they need to be guided — but the consequences for them are potentially severe.
Rolf Claessen: I have another question for you. You were the instructing attorney — do you think the decision was perfect? What argument that you made was the most underappreciated by the court? And where do you think the judgment got it wrong, or was it all perfect?
Bruce Dearling: No, it got 90% or 95% correct. The intermediate step is right. That’s the most important thing in the decision — it’s the intermediate step. The any-hardware thing — that’s logical, that makes some sense — but if people say “if the any-hardware rule is the important bit,” no it isn’t. It’s the intermediate step. That’s the important thing.
Where do they go wrong? I think they went wrong because — and you’ve got to bear in mind that unlike German courts, I’ve got to be careful about how I express this — generally, as I understand it, and correct me if I’m wrong, but the judiciary in Germany on patent cases are generally more technically able. They’re normally technically qualified.
I look at the Supreme Court justices and the Court of Appeal justices — we had one who was a humanities undergrad, one was a chemist. Good luck with trying to argue complex artificial neural network technologies, which are difficult even for me to understand. And I’ve been working in the field. They’re hard to understand. They require real understanding, real appreciation. They could say, well, actually we don’t need to look at the technology — but frankly, if you’re looking at the statutes and exclusions to patentability and asking what a computer program is, then you need to understand what these technical terms really are. And if you can’t, then the judgment is potentially flawed. Their finding that the neural network is a computer program is, I think, technically obtuse.
You know that the Singaporean government — the Intellectual Property Office of Singapore — released about six weeks ago a consultation note to the Singaporean profession and population, asking: is the Emotional Perception case right, and do we need to adopt it into Singaporean national law? So this is direct soft power from the UK Supreme Court changing Commonwealth legislation and statutes. We’ll see what happens. But from what I’ve seen of a draft response from the attorneys, they’re saying essentially: we agree any hardware is right, the intermediate step is right. The assessment of the neural network as a computer program is wrong, or it just doesn’t make any sense. And I’ve made the same comments before in SIPA, in the relevant round in March.
There’s a disconnect. I mean, it’s like they equate a computer program with being able to be run on an analog computer. Now, an analog computer has no central processing unit. An analog computer just has resistors and transistors and capacitors. So if they’re saying that an analog computer can run a program — that’s essentially what they’re saying in part of the judgment. Where is the program in an analog computer? And if they’re saying it’s in the values of the resistors and the capacitors, then that has implications for any circuit we’ve got — it’s potentially a computer program — which is just madness, because it doesn’t sit well with the legislation and decisions we’ve looked at over the last 50 years.
This is a real problem. It may be a storm in a teacup because you can overcome the objections by having any hardware, but it’s an argument they shouldn’t have been making. It seems to be abstract legal argumentation which has little credibility in my personal view, although it’s now law. It may be that someone can take that, have an argument with the Supreme Court, get them to fix this. The other thing is the EPO looks at a neural network as a mathematical method, and the UK now says it’s a computer program. Neither is right.
The EPO is wrong as well. If you look at the actual decision which they regularly quote — the Vicom case — if you actually read the claim and look at the case, you see that it doesn’t make a huge amount of sense. A neural network has applied mathematics in it. It can be based on a computer program because it’s required to set up the learning objectives and the loss function. Mathematical processes — it tweaks the weighting factors of neurons over the course of the training epochs. But at the end of the day, if the function performed by the neural network is new and it’s directed towards a technical implementation which is technically relevant, then it shouldn’t fail for being a mathematical method.
And I think the EPO guidelines actually say that. Even recommendations — the UK court said that a recommendation is not technical. Well, actually it is, because it’s data processing, and you’ve got to work out how does the data processing work to provide an improved recommendation? Again, it goes back to the T 1249/22 decision. There’s a whole raft of these things which are left not entirely resolved. There’s enough here to keep someone busy for a few more years.
Rolf Claessen: Right. So I have a question for you now that we’ve talked about the decision of the UK Supreme Court and the UPC — the Unified Patent Court — with very, very similar wording. What do you say are the three most important takeaways for patent practitioners in the US, in Europe, in the UK, before the EPO? Are there any things that you really want patent practitioners to take away from our discussion here?
Bruce Dearling: Yeah, okay. So first: make sure the claim has some structure in it. You need to have any hardware. That’s number one — in terms of claim drafting.
In terms of the description, you really have to understand what the invention is about. And you’ve got to make sure that you explain what function is achieved by what piece of hardware, kit or software. And if you do that — don’t nickel-and-dime this by writing the claim first — I would suggest that you run into problems. You need to understand what the invention is about. And you need to make sure that the description is complete and full to describe the functionality and the effects that are achieved in the real world. And if you can do that, then you’re on a much sounder basis — much, much stronger. There’s a much stronger foundation for this. So that’s two things. Is there a third one? That’s me being a bit cheeky, but I suppose I know what’s going on.
Rolf Claessen: Yeah, but maybe the third takeaway is that maybe the EPO will rethink the way — at least how AI inventions are assessed for inventive step.
Bruce Dearling: Well, as I said to you before, it could be that that’s the case. I don’t want to repeat myself again. The word “permissive” was used in a conversation I had with respect to the UK Supreme Court approach. COMVIK fundamentally still breaks with me and has done for years, because the way it’s set up and the way it’s applied distorts fundamentally what the invention is about. And until such time as that distortion is removed, there is a problem of objectivity versus subjectivity. And I think that’s really what the EPO has to grapple with. It’s not an easy thing to deal with, but maybe there are things going on.
Bruce Dearling: It’s not an easy thing to deal with. I don’t know who’s going to argue it. It would have been useful for me to still have the original case up and running at the EPO because these arguments would have been fleshed out. I’m pretty sure they would have been referred to the Enlarged Board. We would have got it resolved. So it’s whether or not I can now work this into the existing case to try and get the examining division to — well, they will refuse, I suspect. And then it’ll go to the TBA. And then the TBA will have to look at this, hopefully with the referrals to the Enlarged Board. And then that fixes the problem on a national and international basis.
Rolf Claessen: Yeah. Let’s see.
[Laughs]
Bruce Dearling: No, we don’t know. I mean, you might have a different view. What do you think? Do you think COMVIK is fundamentally right or fundamentally wrong?
Rolf Claessen: Well, I’m not so much into AI inventions. I’m a chemist and I usually deal with chemistry inventions. But from the discussion that we had, I think that the EPO might rethink their position. I don’t know. Let’s see. Let’s hope so.
Bruce Dearling: Well, they liked it. They liked problem-solution. It’s been with us for 25 years. It suggests that it’s a compromise. It’s not mandated by the European Patent Convention — that’s the point. It’s something they think works. And these things only work until such time as someone comes along and says, actually, you’re wrong, and this is the reason.
Rolf Claessen: Let’s see if they choose a different route at least for AI inventions. So Bruce, thank you very much for your insight and for talking about the case that you were involved in with the UK Supreme Court. Where could people reach you if they have more questions about this field — basically patents, AI protection in the UK and Europe — and if they want to ask you more questions about this case?
Bruce Dearling: Sure. Through the Hepworth Brown website or my LinkedIn profile, I suppose. The Hepworth Brown website has an email link. I’m trying to post things on it as well to try and provide a bit more context. But if people have fundamental questions on this stuff, then I’m happy to try and answer them. I suppose that I can be considered to be quite knowledgeable in the area.
Rolf Claessen: Right. Certainly more than I am.
[Laughing]
Bruce Dearling: So I was fortunate. As a consequence of the work I’m doing, I was appointed last year to the WIPO Standing Committee on Patents and Privacy. That was discussed for the issues of where WIPO goes and what the direction of the problems are that we have in high-tech areas. So there seems to be some degree of understanding that I might know what I’m talking about. I think I probably do.
Rolf Claessen: Thank you, Bruce. Thank you very much for being on IP Fridays.
Bruce Dearling: My pleasure. Thank you very much, Rolf.