In the latest UK decision on the patentability of AI inventions, the Court of Appeal has overturned the eye-catching High Court decision relating to Emotional Perception AI Limited's patent application for an AI system which provided improved file recommendations. The Court of Appeal held that an artificial neural network (ANN) for such a system, whether implemented by hardware or software, was a computer. It also applied the test for determining whether a computer program was excluded from patentability, concluding that the Emotional Perception AI system did not make a technical contribution and was therefore excluded. In doing so, the Court of Appeal signalled its agreement with the UKIPO's Hearing Officer's initial decision in BL O/542/22 that the Emotional Perception AI patent application claimed subject matter excluded from patentability. As a result and subject to any further appeal, the brief period of time during which UKIPO examiners have been advised not to reject patent applications for ANNs as subject matter excluded from patentability, now comes to a swift end. On 25 July the UKIPO published revised practice guidance confirming that following the Court of Appeal decision, patent examiners should treat ANN-implemented inventions like any other computer implemented invention.

The decision of the Court of Appeal is unlikely to come as a surprise to most, given the reaction to the High court decision which it overturns, but it will no doubt be seen as a disappointment to those hoping for the UK to become a patent-friendly jurisdiction for inventions relating to ANNs.  The reasoning does, however, set out detailed guidance on the approach that should be taken when assessing patentability of inventions in this field, and the decision appears to be consistent with the assessment of ‘technical contribution’ developed through earlier case law such as Aerotel applying to other computer implemented inventions.

The background

To recap, as discussed in our previous blog post on the High Court decision, Emotional Perception sought to patent a method of providing semantically relevant file recommendations, using an ANN. Much of claim 1 of the patent application defines detail of the training of the ANN to perceive semantic similarity or dissimilarity between media files, and using the trained ANN to recommend a file which is semantically similar to a given input.  The application was refused by the UKIPO as being a program for a computer program “as such”, an exclusion from protection under s 1(2)(c) of the Patents Act 1977.

The refusal was appealed to the High Court which, on 21 November 2023, issued a contrary decision that the computer program exclusion was not, in fact, invoked. The idea of using pairs of files for training and setting parameters for programming the ANN accordingly was considered to be a concept extending further than the programming itself, so that the claim went beyond a computer program “as such” and was not excluded from patentability.  

The approach of the Court of Appeal

LJ Birss went back to basics to determine the nature of a computer, and that of a computer program. Defining a computer as a “machine for processing information”, an ANN was considered as “clearly a computer”.  Defining a computer program, as “a set of instructions for a computer to do something”, the weights of the ANN, applied to each input to a neuron in the network, were considered as a computer program since a particular “set of weights will cause the machine to process information in a different way”. No distinction was seen between the weights of an ANN and logical ‘if-then’ statements associated with more traditional computer programs.

The Court of Appeal concluded that since the training of an ANN was therefore part of the creation of the ANN itself, and thus therefore the creation of a computer program, it was not of significance for the purpose of assessing the patentability of the program itself. For this purpose, it was the contribution of the program which was to be considered, with the training of the program, being ‘subsidiary in nature and irrelevant’.  In this case, the contribution was providing improved file recommendations. but this was not seen as a technical improvement to the running of a computer. LJ Birss stated “It is the semantic similarity of the files here which gives rise to their recommendation but that is not a technical matter at all". 

The technical contribution of a computer program, identified using the five signposts set out in AT&T Knowledge Ventures, and endorsed in HTC v Apple are thus as relevant to ANNs as any other computer program:

1. whether the claimed technical effect has a technical effect on a process which is carried on outside the computer;
2. whether the claimed technical effect operates at the level of the architecture of the computer; that is to say whether the effect is produced irrespective of the data being processed or the applications being run;
3. whether the claimed technical effect results in the computer being made to operate in a new way;
4. whether the program makes the computer a better computer in the sense of running more efficiently and effectively as a computer;
5. whether the perceived problem is overcome by the claimed invention as opposed to merely being circumvented.

None of these signposts was considered to point towards a technical contribution of the Emotional Perception computer program, such that there no grounds on which a patent could be granted.

Key takeaway

In summary, the Court of Appeal decision will likely be welcomed by those seeking consistency of approach in UK case law, who may have raised eyebrows at the High Court decision. The new dawn for applicants in this area has, however, been short-lived.

LJ Birss notes, however, that “the fact that s1(2) of the Act is engaged in a case of an ANN implemented invention, as much as it would be in any computer implemented invention, does not mean it is unpatentable….ANN implemented inventions are in no better and no worse position than other computer implemented inventions”. The consistency between an ANN and a computer-implemented invention is something which was also seen in the European Patent Office Decision T 702/20, in which “a neural network relates to both programs for computers and to mathematical methods”.

The decision is therefore not one which should cause any disturbance to the wider issue of how to assess patentability of software in general, and simply incorporates the weights of an ANN within the meaning of “computer program, as such”. Therefore, neural networks can in fact be patented under both jurisdictions, subject to the usual considerations relating to the presence of a technical contribution or effect, something which was not present in this case.