Case Law Opinion – Emotional Perception AI Limited v Comptroller General of Patents
In the UK, patent protection for computer software has been a controversial and difficult topic for a long time. Perhaps surprisingly, it has taken nearly 50 years since the UK Patents Act 1977 (implementing the then-new European Patent Convention) came into force for a case about whether a computer program can be patented to come to the UK Supreme Court.
Less surprisingly, given the fuss being made about artificial intelligence at the moment, the invention at issue in Emotional Perception AI Limited v Comptroller General of Patents[1] is about an artificial neural network. And given the amount of money at stake now in getting people to engage with online content, it probably isn’t a particular surprise either that this invention is about providing better recommendations (along the lines of “if you liked that, try this”), for example in an online music streaming service.
Actually, what the invention is exactly turns out not to be terribly important for the main breakthrough now made by the Supreme Court. But there seems to be no room to criticise the judges’ understanding of the technology here – the invention, and artificial neural networks generally, are concisely but accurately explained in the introduction to the judgment.
The Court hasn’t actually decided whether this particular invention is patentable or not. That will need to be the subject of further examination by the UKIPO, and it isn’t beyond the realms of possibility that it will come back to court again, maybe even the Supreme Court again. So what have the Supreme Court actually said, and why is it such a big deal?
Well, the statute says that, among other things, programs for computers “shall not be regarded as inventions”. But it also says that they are only excluded to the extent that the patent application relates to such subject matter (computer programs, in this case) “as such”. For a long time now the UK Courts, and the UKIPO, have assessed whether this exclusion bites according to what is known as the Aerotel test, after a Court of Appeal judgment in 2006[2]. The Aerotel test is not, however, what has been practised at the European Patent Office. Whether it boils down to the same thing in the end has been endlessly debated, but now it doesn’t matter, because what the Supreme Court have said is that we should forget about Aerotel and do what the EPO have been doing for years.
So how do we now interpret this exclusion from patentability?
Patents are for protecting inventions. The EPC and the Patents Act say that. But the word “invention” is used here in quite a precise way, not necessarily how you might immediately think. In particular, the law goes on to set out the criteria by which an invention shall be considered to be new, or not, and by which an invention shall be considered as involving an inventive step (it is, if it is not obvious). So what that means is that the law envisages you can have things which are not inventions at all and so can’t be protected by patents (they are not “patent eligible” to use the term favoured by Americans). Then, there are inventions which turn out not to be novel, and inventions which while novel can’t be patented because they are obvious. Once you have something which is an invention which is novel and also non-obvious, you can have a patent (subject to one or two other exclusions of minor relevance here).
A step-by-step approach is therefore called for. If the claim really only has excluded features, then it falls at the first hurdle – it isn’t eligible, if you like. After that, it is assessed for novelty and inventive step. Assessment for novelty and inventive step is perfectly routine, but we just have to be careful that what we are assessing is the invention, which might not be made up of every single claimed feature – features which contribute nothing to the technical effect of the invention don’t count. So there has to be an “intermediate step” to make sure that we are assessing an actual invention, and not a whole load of non-technical features cunningly disguised as one. Both technical and non-technical features might “count”, but only if they contribute to the invention, which can’t be something like a computer program which is excluded as such.
This interpretation of things isn’t new – the EPO worked it out years ago. But if the UK courts have erred before then they seem to have redeemed themselves here with a concise and convincing argument, which only makes you wonder why anyone ever thought differently. I hope I have done it some justice in this short summary.
So what about Emotional Perception’s application? Well, having said that the UKIPO and the lower courts had just got themselves in a muddle trying to apply this Aerotel test, the Supreme Court didn’t spend any time on the invention itself. It was beyond argument that the claim includes some technical features and there is therefore an invention, but whether that invention is novel or inventive remains to be seen. The “intermediate step” will have to be carried out, and it seems to me a long way from certain that the result will be a patent in the end. What the Supreme Court have done, for this and other future cases that involve software, is to remove an enormous distraction. You can still argue over whether Aerotel gets to the same result as the EPO’s (and now the UK’s) approach, if you like. But the argument is now clearly pointless. The battleground that matters will remain over what counts as “technical”, and that is why we are not yet at the end of this interesting story.
We are lucky, I think, that we have eventually arrived here. Removing the distraction of Aerotel is a valuable result, but the proceedings were almost derailed. In the High Court, apparently as a result of what was understood to be a concession made by UKIPO’s counsel, the Judge reached the frankly bizarre conclusion that artificial neural networks were just not computer programs at all. This was put right by the Court of Appeal, and the Supreme Court doesn’t seem to have wasted too much time on it. But confusion still reigns – I have seen several reports that the Supreme Court has ruled that “artificial neural networks can be patented” or are “no longer automatically excluded”. Actually I’ve had dozens of patents granted for ANN-based inventions before and during the Emotional Perception litigation, many of them with little fuss. Artificial neural networks are treated just like other software inventions – showing a technical effect is key. In many cases we can be fairly clear whether there is a technical effect or not, but there is still a battleground at the borderline. For my part I think that Emotional Perception’s invention is probably not patentable based on the current body of law on “what is technical”, but it may be close enough to the borderline to have a good push, and might make for more genuinely interesting case law.
So will all this make it any easier to get patents for computer-implemented inventions? I tend to be of the view that not much about the substance of the law has changed here – Aerotel wasn’t wrong, just misunderstood. But if what replaces it is easier to understand, then that can only lead to less confusion and more consistent results for would-be patentees. No, all software hasn’t been made patentable overnight, but we’ve got a distraction out of the way which was making this difficult question even harder than it needed to be.
[1] [2026] UKSC 3
[2] Aerotel Ltd v Telco Holdings / Macrossan’s Application [2006] EWCA Civ 1371
Author
