In recent years, the UK Intellectual Property Office (UK IPO) attitude towards the patenting of AI inventions has been fairly strict. However, a High Court decision handed down last week has adopted a much more permissive approach which puts the UK out of step with the practice of the European Patent Office (EPO).

The High Court decision ([2023] EWHC 2948 (Ch)) arises from an appeal by Emotional Perception AI Ltd against a UK IPO decision to refuse its patent application under section 1(2) of the UK Patents Act (excluded fields).

The invention relates to training a neural network to perceive semantic similarity or dissimilarity between media files such as music files, and then using the trained neural network to recommend a file which is semantically similar to a given input, including sending the file to a user device over a communications network.

The neural network is trained by comparing two different distances between two data files. One of the distances is derived using natural language processing of the semantic text description of the respective file, while the other distance is obtained from the neural network using "objectively discernible properties" (e.g. timbre, rhythm, and tonality). The training goal is to minimise the difference between these two distances.

At the UK IPO, the hearing officer found the subject matter of the claims to be excluded as a computer program as such, finding that providing a user with a better recommendation, such as a song they are likely to enjoy, is subjective and not technical. This finding accords with EPO practice on recommendation tasks (see 5.2 of T 306/10).

The High Court decision reverses the UK IPO decision, finding that the neural network operates at a different level from the underlying software on the computer and so is not a computer program.

Perhaps even more significantly, the decision finds that even if the claim were to a computer program, there would be a technical contribution such that the invention avoids the exclusion. In reaching this view, the judge took into account that the method involved moving data outside of the computer system in the form of a file that is transferred to a remote user device. The judge also took into account the method of selection, which was found to involve identifying the file as being semantically similar to the input file "by the application of technical criteria which the system has worked out for itself". It was not a bar to patentability that the benefits of the invention might be said to lie in the subjective effects of the output for the user.

Much of the reasoning behind the decision relies on a comparison with a hardware neural network. In particular, there is a suggestion that a trained neural network produced by a training process may itself constitute an external technical effect. The point is made in relation to a hardware ANN (artificial neural network), but the judgment indicates that there ought to be no difference between a hardware ANN and an emulated ANN for these purposes.


Patents in the UK are obtainable both through the UK IPO and the EPO, with any subsequent enforcement/UK validity proceedings taking place before the UK courts. While there have always been some differences between the UK and EPO approaches, the UK courts have previously argued that the UK approach to eligibility is reconcilable with that of the EPO and ought to reach the same outcome (at least usually). The decision puts that suggestion under significant strain.

The UK IPO is bound by decisions of the UK courts and has recognised the significance of the case already by suspending the applicability of its previous guidance for AI inventions.

The UK IPO has the option to appeal the decision. If it decides to do so, all eyes will turn to how the Court of Appeal approaches the issue: will it undo the judgment and favour harmonisation, or use this as an opportunity perhaps to recognise AI as a new technology area that ought to be better supported by the UK patent system? In that case (or in the case of no appeal), the UK may be on a collision course with the current attitude of the EPO. Such a course could have profound consequences in the UK and across Europe for incentives to invest and innovate in the future of AI.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.