Hepworth Browne

What the Emotional Perception AI Supreme Court Decision Means for UK Patent Applicants

On 11 February 2026 the UK Supreme Court handed down its judgment in Emotional Perception AI Limited v Comptroller General of Patents. The unanimous decision fundamentally reshaped how UK patent law treats AI and computer implemented inventions.

Published9 June 2026
TopicAI
FieldArtificial intelligence, IPC G06N
Reading time11 minutes

On 11 February 2026 the UK Supreme Court handed down its judgment in Emotional Perception AI Limited v Comptroller General of Patents, Designs and Trade Marks [2026] UKSC 3. The five justice panel, Lord Briggs, Lord Hamblen, Lord Leggatt, Lord Stephens and Lord Kitchin, issued a unanimous decision that fundamentally reshaped how UK patent law treats artificial intelligence, artificial neural networks and computer implemented inventions. It has wider implications for other technologies as well.

For businesses developing AI systems, software led products and neural network based technologies, this is a landmark ruling that changes the way patentability objections must now be analysed in the UK. For patent attorneys, examiners and applicants who have spent years working within the framework laid down by the Court of Appeal in Aerotel Ltd v Telco Holdings Ltd [2006], that old contribution with rider framework is now gone. Understanding what replaced it, and what that means in practice, is the essential starting point for any AI patent drafting and prosecution strategy in the UK.

What was the Emotional Perception AI case about?

The case concerned a patent application for a system and method that uses an artificial neural network to generate file recommendations. The technology trains an ANN to replicate human emotional response to perceived similarity between media files, such as music tracks, videos or text, determined merely by reference to objectively measurable physical properties of those files rather than their semantic descriptions.

The invention was designed to improve recommendation services and content discovery search tools. At the heart of the application was a method of training an ANN to converge and ultimately align separation distances between file pairings in a physical property embedding space with corresponding separation distances in a semantic embedding space, the latter derived from how humans describe the same files in natural language. Once trained, the ANN could make semantically meaningful recommendations using only physical file properties, with the mapping retaining semantic information that is otherwise lost.

The UK Intellectual Property Office originally rejected the application on the basis that the claimed invention was excluded, under section 1(2)(c) of the Patents Act 1977, from patentability for being a program for a computer as such. Section 1(2) has, according to the statute, as near as practicable the same effect as article 52(2)(c) of the European Patent Convention. The High Court allowed an appeal, the Court of Appeal reversed that decision and restored the rejection, and the case then reached the Supreme Court on a further appeal. That appeal thrashed out the fundamental issues surrounding the tenets of patentability, namely the treatment of invention and how inventive step and obviousness are assessed.

Hepworth Browne acted as the instructing firm throughout these proceedings, with Bruce Dearling the case architect and instructing patent attorney for Emotional Perception AI. That direct involvement gives the firm detailed insight into the arguments developed across every level of the litigation and the practical consequences of the ruling.

What did the Supreme Court decide?

The court resolved three principal issues and, in doing so, produced a ruling with far reaching consequences for UK patent practice and, indeed, implications on a transcontinental basis.

Issue 1: should the Aerotel approach be abandoned?

Yes. The court held that the four step Aerotel approach, applied by the UKIPO and UK courts for nearly two decades, should no longer be followed. It was unsound. The court accepted that Aerotel was inconsistent with the approach approved by the Enlarged Board of Appeal in G1/19 Bentley Systems, which had firmly endorsed the any hardware approach as the correct method for determining whether a claim amounts to an invention within the meaning of Art.52(2) EPC.

The Board of Appeal had for years criticised Aerotel as not being a good faith implementation of article 52, the decision being incompatible with it and based on a misunderstanding of the meaning of invention as used in the EPC. The Supreme Court accepted those criticisms and held that the approach, which asked what the inventor had really added to human knowledge as a way of testing whether the contribution fell within excluded subject matter, was methodologically flawed. It jumbled together the question of what constitutes an invention with the separate and independent requirements of novelty and inventive step, which must be assessed at a later stage.

In its place the court confirmed that the UK should adopt the same interpretation of Art.52(2) and (3) EPC, namely the any hardware approach. Under that approach a claim involving any piece of hardware, including a conventional computer, a computer readable storage medium or other technical means, is not a claim to excluded subject matter as such and therefore qualifies as an invention for the purpose of article 52.

Issue 2: is an ANN a program for a computer?

Under UK law the answer is presented as yes. The court rejected the argument that ANNs are not programs for computers at all and therefore fall entirely outside the exclusion. It held that an ANN, whether implemented in dedicated hardware or emulated in software on a conventional computer, constitutes a set of instructions to manipulate data in a particular way so as to produce a desired result, and is therefore a program for a computer within the meaning of Art.52(2)(c) EPC. The distinction between a hardware ANN and a software emulation was rejected as conceptually untenable. That assessment is inconsistent with the approach of the EPO, and it is arguably wrong, based on abstracted legal logic and a somewhat obtuse technical argument by the court.

Issue 3: was the claimed invention excluded as a computer program as such?

No. Applying the any hardware approach, the court held that the claimed invention is not excluded because the claim had hardware elements. The ANN clearly requires implementation on some form of computer hardware, and the claims also referred to a database, a communications network and a user device. The presence of hardware is sufficient to bring the invention within the meaning of invention under Art.52 EPC and section 1(2), and to take it outside the as such exclusion. The application was therefore wrongly refused on those grounds, and the matter was remitted to the UKIPO to assess novelty and inventive step.

What is the any hardware approach and why does it matter?

The any hardware approach is a fundamentally lower threshold than the one applied under Aerotel. Under the old test the analysis focused on what the inventor had added to human knowledge and required that, in addition, the contribution in totality was technical in nature. That consistently produced problematic outcomes for both AI and software applicants, because it made it easy for examiners to characterise the real contribution as arbitrarily excluded without a proper understanding of the nature of invention. It is a backward looking approach, and it had implications for other fields of activity including business methods.

A method for performing a mental act, a mathematical method and a program for a computer, all as such, are statutory exclusions under both Art.52(2) and (3) EPC and section 1(2). However, without adequately considering the technical means through which the contribution was achieved, Aerotel was fundamentally flawed.

Under the any hardware approach, the exclusion hurdle is cleared by the mere presence of hardware.

Emotional Perception AI v Comptroller General [2026] UKSC 3

The real analytical work has now moved to the intermediate step. Whilst that is a filter, it operates to identify those features of the invention that contribute, through interaction and in a causal positive way, to the overall technical effect. Only then can novelty and inventive step be assessed. The court was clear that adopting any hardware does not mean every AI or computer implemented invention will be patentable. It means the question must be addressed properly, at the right stage and with the right framework.

What is the intermediate step?

The intermediate step is one of the most practically significant aspects of the guidance. It arises because the any hardware threshold is deliberately low, and passing it does not mean the invention is automatically patentable. This intermediate stage investigates whether claim features, particularly non technical features, interact within the context of the invention as a whole and therefore contribute to its technical character.

That matters because claims for computer implemented inventions, in all fields of technology, are not infrequently mixed claims including both technical and non technical features. Viewing these in isolation removes the context of the invention. It was recognised by the EPO, at both its Technical Board of Appeal and its Enlarged Board, that non technical features can contribute in a positive sense to patentable innovation. Non technical features that interact with the technical subject matter of the claim and contribute to a technical effect are included in the subsequent inventive step analysis. Those that do not contribute are excluded, which is the line expressed in the Comvik decision T641/00 at headnote 1.

The practical consequence is that applicants and their attorneys need to ensure the technical interactions between claim features are clearly identified and properly articulated. A claim that includes a technically significant interaction between an ANN architecture and other system components needs the specification to describe that interaction in a way that makes the real world effect evident. Getting this right at the drafting stage is significantly easier than reconstructing the argument later in prosecution, not least because of procedural objections relating to added matter.

Does the decision affect inventive step analysis?

Yes, but not by requiring UK courts to adopt the problem and solution approach. The court was explicit that the Pozzoli test remains a legitimate approach in the UK and that there is no requirement to change to the EPO methodology. Rejecting Aerotel does not require wholesale adoption of EPO practice, only alignment on the interpretation of Art.52 EPC.

The modification is the intermediate step. Features that are non technical but which interact, in the context of the invention as a whole, with the technical subject matter of the claim can support the presence of an inventive step, provided the totality of the claimed invention is not obvious over the prior art. For AI applicants this means careful thought at drafting and in argument about which features are relied upon, and whether they genuinely interact with the technical subject matter to deliver an effective technical effect.

What should AI applicants do now?

The immediate implication is to reassess UK filing strategy in light of the new position. Applicants with pending UK applications that met exclusion objections under the old framework should consider whether those objections can now be overcome with the express inclusion of any hardware. Applicants with refused or abandoned applications declined primarily on exclusion grounds, rather than novelty or inventive step, may have grounds to revisit their position, although that may require appeal through the courts. The decision and strategy is determined by the precise state of the file, and that may still not be finally closed.

For new filings the decision creates a stronger basis for framing AI related claims around the technical interactions between system components, the contribution of specific architectural choices to technical character, and the way trained model behaviour contributes to a technical effect. Claim drafting that clearly identifies and preserves those interactions is likely to be more resilient under the intermediate step analysis.

Frequently asked questions

Is Aerotel still the law in the UK? No. The Supreme Court has held that the approach should not be followed. It has been replaced by the any hardware approach endorsed in G1/19, combined with the intermediate step analysis derived from the principles in T154/04 Duns Licensing.

Does the ruling mean AI inventions are automatically patentable in the UK? No. It depends heavily on the overall technical effect and how it is expressed. The judgment clearly lowers the exclusion hurdle, but patentability still depends on novelty, inventive step and the proper identification of features that contribute to technical character. Not every AI system or software product will satisfy those requirements, especially if it is effectively the automation of an existing process.

Who acted in the case? Emotional Perception AI was represented by Mark Chacksfield KC and Edmund Eustace as counsel, instructed by Hepworth Browne. Bruce Dearling was the instructing patent attorney and was responsible for the fundamental arguments and legal framework from inception through to the Supreme Court decision.

Does the ruling affect non AI software and computer implemented inventions? Yes. It changes the exclusion analysis for all computer implemented inventions and, moreover, all fields of technology as required by the TRIPS agreement. Any applicant with software related claims affected by the old framework may need to review their position.

What happens to the application now? The Supreme Court allowed the appeal, setting aside the Hearing Officer decision and remitting the application to the UKIPO for assessment of novelty and inventive step. Whether the invention satisfies those requirements has not yet been determined.

For detailed guidance, explore our artificial intelligence and computer programs practice, or contact the firm to discuss a specific application or portfolio.

Get in touch

Tell us what you have made.

A first conversation costs nothing and usually saves money. Send us an outline and we will tell you honestly whether there is something worth protecting, and what it is likely to cost.

Step 1 of 4

Where you are

What stage are you at?

This tells us who should pick up your enquiry and how quickly.

What you need

Which right fits best?

A rough steer is fine. We will confirm the right route once we understand your invention.

The invention

Give us a short outline.

Keep it high level. Please do not send confidential detail before we confirm terms.

Where to reply

Who shall we get back to?

Please add your name
Please add a valid email address

Enquiry ready

Thank you. An attorney will review your outline and respond within one working day.