Getting technical: Lessons from the emotional perception Remittal

The UK Supreme Court's decision in Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2026] UKSC 3 has changed the way in which excluded subject matter is approached in the UK. However, following the case's remittal to the UK Intellectual Property Office (UKIPO), the patent application at the centre of that decision has now been refused again.

The remittal decision provides an important indication of what the Supreme Court's new approach is likely to mean in practice. In particular, just because an invention involves technical means does not mean that all its features will contribute to inventive step. Consistent with established practice at the European Patent Office (EPO), the more important question is which features contribute to a technical solution to a technical problem.

A new approach to excluded matter

The application concerned an artificial neural network (ANN) used to provide semantically relevant file recommendations. Broadly, the ANN is trained so that similarity between files in terms of objectively measurable properties corresponds more closely to their similarity in terms of semantic descriptions. Once trained, the ANN can be used to identify semantically similar files based on their measurable properties.

The Supreme Court concluded that the Aerotel approach previously applied in the UK should no longer be followed. Instead, the Court adopted an approach based on the principles set out in the EPO’s Enlarged Board of Appeal decision G 1/19. This represents a significant departure from established UK practice that emerged after Aerotel. Although the Aerotel approach may have been intended to produce broadly consistent outcomes with the EPO approach, in practice the two did not always lead to the same result. The Supreme Court's decision brings the UK approach to assessing excluded subject matter much closer to that of the EPO.  

Under the new approach, the initial threshold for establishing that claimed subject matter is an "invention" is relatively low. G 1/19 adopts the "any hardware" approach, under which reference to a computer or other technical means can be sufficient to clear the Article 52 EPC exclusion.

The Supreme Court held that Emotional Perception's claims passed that threshold. Although it concluded that an ANN is a "program for a computer", the claimed subject matter was not a computer program "as such". The ANN had to be implemented on some form of computer hardware and the claims expressly referred to hardware including a database, communications network and user device. This was sufficient for the claimed subject matter to have technical character.

That, however, was only the first hurdle.

The Supreme Court identified a further "intermediate step", derived from G 1/19, between determining whether the claimed subject matter constitutes an invention and assessing inventive step. This considers which features contribute to the technical character of the invention as a whole. The UKIPO subsequently set out the resulting three-stage approach in its practice notice, published on 14 July 2026: first, determine whether the claimed subject matter is an invention (i.e. whether it passes the "any hardware" test); second, identify which features contribute to its technical character (this is the "intermediate step"); and third, assess inventive step based on the features making that contribution. As expressed in the practice notice, a feature makes such a contribution where it contributes to, or interacts with, the technical subject matter of the claim to provide a technical solution to a technical problem. Features which do not do so cannot support inventive step.

Importantly, this is not simply a question of dividing a claim into features which appear technical and features which appear non-technical. The Supreme Court explained that even a feature which is technical in isolation may fail to contribute if it has no technical function in the context of the claimed invention. Conversely, a feature which is non-technical in isolation may nevertheless contribute where it interacts with the technical subject matter of the claim in solving a technical problem.

The Supreme Court did not itself decide how that analysis should apply to Emotional Perception's claims. It instead remitted the application to the UKIPO, observing that it had not received sufficient argument on how the new intermediate step should be applied to the particular invention.

Applying the new approach

The remittal was not the UKIPO’s first opportunity to apply the new approach. An earlier Hearing Officer decision BL O/0673/26 in relation to an application by FLIR Belgium BV, issued in July 2026, had already considered which claimed features contributed to the technical character of an invention when assessing inventive step. The Emotional Perception remittal is nonetheless particularly illustrative because it shows how the new approach was applied to the application at the centre of the Supreme Court proceedings themselves.  

On 1 October 2026, the UKIPO issued its decision following remittal in BL O/0894/26. The application was refused again, but on different grounds from the initial refusal. Following the Supreme Court judgment, the application was examined under the new approach and the outstanding questions were considered by the Hearing Officer.

The Hearing Officer focused on claim 4, a computer-implemented method of providing semantically relevant file recommendations. In broad terms, the claimed method trained the ANN so that distances derived from measurable properties of files converged towards corresponding semantic distances. The trained ANN was then used to identify and output files considered semantically similar to a target file.

Applying the new intermediate step, the critical issue was which features contributed to the technical character of the invention and could therefore be taken into account when assessing inventive step. The Hearing Officer concluded that outputting a "semantically better" file was not a technical result. Consequently, the ANN's learning objectives and its training to close the "semantic gap" could not contribute to the technical character of the claim and could not therefore support inventive step.

Once those features were disregarded for that purpose, what remained (specifically, “Computer hardware running a program that selects and outputs a file”) was considered obvious over the common general knowledge. The application was therefore refused for lack of inventive step.

What are the key questions?  

This new approach puts considerable emphasis on a question already central to the patentability of computer-implemented inventions before the EPO: is there a technical solution to a technical problem, and which features provide it?  

This is a question I regularly ask inventors working in areas close to the boundary of patentability. It can provide a useful early indication of whether patent protection is likely to be appropriate, particularly for inventions where the innovation appears to lie in subject matter that the EPO may consider non-technical. Common examples include classifying documents solely according to their textual content, automating a purely business or administrative process, or presenting information to a user based only on its cognitive content.

Perhaps more importantly, this question encourages consideration of where the technical innovation actually lies. For example, despite a product’s apparently non-technical purpose, is there nonetheless a non-obvious technical solution to a technical problem in its implementation? If so, patent protection may be more likely to be available.

What’s really changed?  

The impact of Emotional Perception is more nuanced than simply expanding the circumstances in which software and AI inventions can be patented in the UK. Clearing the relatively low initial “any hardware” hurdle still leaves the often more difficult question of which features of the invention actually contribute to its technical character and can therefore support inventive step.

Precisely how the new three-stage approach will be applied in practice is likely to develop as the UKIPO gains experience of it and further cases are decided, including through any appeal of the present decision. The earlier FLIR decision already illustrates the importance, and potentially fact-sensitive nature, of determining which features contribute to the technical character of an invention. It may therefore take some time for outcomes to become more predictable.  

For applicants considering protection for AI and other computer-implemented inventions in the UK, the underlying question is as important as ever:

Which features contribute to a technical solution to a technical problem?  

The renewed refusal of Emotional Perception's application is a timely reminder of the importance of asking this question from the very start.

References

Emotional Perception AI Limited v Comptroller General of Patents, Designs and Trade Marks [2026] UKSC 3, UK Supreme Court, judgment of 11 February 2026. Full judgment (PDF)  

Search and Examination of UK Patent Applications under the Patents Act 1977 (as amended), UKIPO practice notice, published 14 July 2026. UKIPO guidance on GOV.UK  

Emotional Perception AI Ltd, BL O/0894/26, UK Intellectual Property Office, decision of 1 October 2026. Full decision (PDF)  

FLIR Belgium BV, BL O/0673/26, UK Intellectual Property Office, decision of 29 July 2026. Full decision (PDF)
‍

Developing an AI or software innovation? Understanding what makes your invention technically distinctive can be critical to securing patent protection.

Speak to author Felix Hall or contact us about protecting your innovation.