Patentability of Artificial Neural Networks to be decided by UK Supreme Court on Wednesday 11 February

On Wednesday 11 February the UK Supreme Court will hand down its much-anticipated judgment in Emotional Perception AI v Comptroller General of Patents.

To get you up to speed, this article covers what's happened so far and the key issues the Supreme Court will be deciding.

Background

Emotional Perception AI applied for a UK patent claiming a process for using an Artificial Neural Network ("ANN") for associating musical tracks together based on both their emotional and musical similarity, and using this to recommend a musical track to the user based on this similarity (for further information on this, see our earlier article here).

The application was rejected by the UK Intellectual Property Office, citing the exclusion from protection in the Patents Act 1977 for "a program for a computer … as such" (section 1(2)(c)).

Emotional Perception AI appealed to the High Court. The appeal raised two core questions:

1. Is an ANN a program for a computer?
2. If it is, does Emotional Perception AI's application nonetheless make a technical contribution allowing it to fall outside the exclusion to patentability?

The High Court: An ANN is Not a Computer Program

The High Court judgment focused firstly on the method by which ANNs are created. Traditional computer programs are generally reducible to a series of "if … then" statements designed by a programmer. In contrast, an ANN is developed using a learning approach where a set of model weights is iteratively adjusted based on exposure to training data.

The High Court concluded that the claimed ANN could not be considered a computer program and was not subject to the usual exclusion to patentability; what was claimed was a set of trained weights for an ANN which had resulted from this "learning" approach and not a traditional computer program.

The High Court also considered whether, if it was wrong about the application of the computer program exception to ANN's, there was a technical contribution through the sending of a file and message in accordance with the recommendation. Here the High Court noted that Gemstar-TV Guide International Inc v Virgin Media Inc found that movement of a file from one storage medium to another could amount to a technical contribution. In the High Court's view, this was a sufficiently analogous situation to how Emotional Perception AI recommended a music track and it concluded there was a technical contribution.

The Court of Appeal: An ANN is still a Computer Program

Both aspects of the High Court's decision were subsequently overturned on appeal.

The Court of Appeal held that an ANN is a computer program because its trained weights and architecture function as a set of machine‑executable instructions, regardless of whether those instructions were written by a human or learned automatically. The Court further rejected distinctions based on the learning process, problem complexity, or permanence of trained parameters, finding these features irrelevant to the statutory exclusion in s.1(2)(c). The ANN was therefore found to be a computer program and excluded unless it made a technical contribution.

The Court of Appeal found that the invention's contribution lay solely in recommending media files based on inferred semantic or emotional similarity, which amounted to a presentation of information and therefore fell within an excluded category. Because any improvement was confined to the quality or aesthetic character of the recommendation rather than a technical process, the ANN provided no technical contribution capable of taking it outside the statutory exclusion.

The Supreme Court

Patentability of computer‑implemented inventions has long been a difficult and relatively under‑litigated area in the UK, and the issues raised by the Emotional Perception appeal go to the foundations of how software—and increasingly AI—should be treated under section 1(2) of the Patents Act 1977. The Supreme Court's decision to hear the case reflects its legal significance: it concerns the proper interpretation of "computer" and "computer program", the correct test for identifying a "technical contribution", and whether the UK should remain aligned with or diverge from European Patent Office (EPO) approaches to AI and computer‑implemented inventions.

If the Supreme Court allows the appeal and revives the High Court's position that trained ANNs differ meaningfully from conventional computer programs, the effects could be far‑reaching. A ruling that expands the scope of patent‑eligible AI implementations would likely encourage applicants to draft and position their AI inventions more aggressively, potentially shifting UK practice in favour of recognising the patentability of machine‑learned functionality.

Conversely, if the appeal is dismissed and the Court of Appeal's reasoning is affirmed, AI inventions will continue to sit within the established software‑exclusion framework, where only genuine technical effects—rather than cognitive, semantic or aesthetic improvements—can confer patentability. This would maintain the existing balance: core technical innovations remain protectable, but improvements in algorithmic recommendation, meaning‑based outputs or other subjective effects would continue to fall to copyright, trade secrets and commercial advantage rather than patents.

A further possibility is that the Court might refine the UK approach to technical contribution—potentially by drawing more closely on the EPO's structured analysis under the "any hardware" and COMVIK frameworks. A recalibration of how technical contribution is assessed in the UK could reshape the treatment of AI inventions without rejecting the premise that ANNs are computer programs. Such an outcome would mark a significant evolution in UK practice and could provide a more predictable patent framework across Europe for rapidly developing AI technologies.