Artificial intelligence

UK High Court finds AI invention patentable

Published on 30th Nov 2023

Court gives guidance for the first time on whether AI inventions are excluded from patentability as computer programs

People in a meeting and close up of a gavel

In a positive move for artificial intelligence (AI) innovators, the High Court has recently overturned a decision from the UK Intellectual Property Office (UKIPO), instead holding that the hearing officer had erred in finding that an AI invention was a computer program "as such", and therefore was not patentable.

In Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks, Sir Anthony Mann found that a system for providing media-file recommendations to end users based on an artificial neural network (ANN) that had been trained in a distinct manner was capable of being patented. It neither falls within the statutory computer program patentability exclusion, he found, nor does it even constitute a computer program at all.

The judge acknowledged that this was the first time the computer program patentability exclusion had been considered in the context of AI. The decision has potentially wide-reaching implications and opens the door to patent protection for many AI inventions. The UKIPO has temporarily suspended the use of its examination guidelines on patent applications relating to AI while it considers the judgment.

The invention

The patent in issue claims an improved system for providing media file recommendations to end users. It does so by training the system on pairs of media files. The ANN analyses files that are accompanied by a natural language description of the semantic properties of the file (that is, a description of how a human would perceive the music). A second ANN analyses the same pairs of files for their physical properties (for example, tone, timbre, speed and so on). The main ANN is then trained to adjust its analysis of the files based on the analysis conducted by the second ANN for the same file pairs.

After repeated training, during which the ANN adjusts its own internal workings, the ANN is able to discern the semantic similarity of a file from its physical properties. Based on this training, the ANN can take a file from an end user, determine its physical properties and make recommendations of semantically similar files from an overall database. It then communicates the recommendation by sending a message and the recommended file to the end user.

The claimed invention was wide enough to cover both a hardware ANN and a computer emulated ANN (that is, an emulation of a hardware ANN run on software).

UKIPO's decision

At the UKIPO, the hearing officer had followed the staged approach and signposts set out in case law to determine whether the invention fell foul of the computer program exclusion. He concluded that the invention was "no more than a computer program" and the invention's contribution was "no more than a computer programming activity". The hearing officer also rejected the contention that the provision of file recommendations was a technical effect over and above the running of a computer program, finding the benefit of the recommendation to be "of a subjective and cognitive nature" instead.

Computer program?

The patent applicant, Emotional Perception AI Ltd, appealed the decision by arguing that the computer program exclusion was not engaged at all as the invention is not a computer program. The judge agreed.

The UKIPO conceded during the appeal that the computer program exclusion would not have applied to a hardware ANN but continued to argue that it applied to an ANN emulated in software. However, the judge stated that the emulated ANN (which was fixed in terms of its operation since it had been trained) was "operating at a different level (albeit metaphorically) from the underlying software on the computer". It was operating in the same way as a hardware ANN and, as such, he held that if the computer program exclusion did not apply to a hardware ANN then it should not apply to an emulated ANN either. Neither type of ANN was held to be a computer program "as such" for the purposes of the exclusion from patentability.

Technical effect?

Given that it had been acknowledged that a computer program was present during the ANN training stage, and in case his finding that the invention was not a computer program was wrong, the judge went on to consider obiter the question of technical effect.

Emotional Perception had argued alternatively that even if the invention was a computer program, it was not a computer program "as such", as the claimed invention had a "technical effect". Emotional Perception argued that the technical effect claimed was the sending of the improved recommendation by message to the end user.

After considering a number of previous cases, the judge stated that the hearing officer had been wrong to find the benefit of the recommendation to be subjective and therefore not technical. Although the criteria by which the file recommendation is selected cannot be described in "purely technical terms", the ANN had made its analysis and selection in a technical way. The judge said that this was sufficient to meet the requirement of a technical effect to avoid the exclusion and, therefore, the invention was not a computer program "as such".

Moving away from considering the technical effect from the perspective of the outcome of the invention (the improved recommendation message), Sir Anthony Mann also went further by noting that a trained hardware or emulated ANN would itself be capable of being an external technical effect. This would prevent the exclusion applying to any computer program used in the training process.

Osborne Clarke comment

This decision is extremely positive for AI innovators, positioning the UK as an AI-developer friendly jurisdiction and confirming that AI inventions can be patented. For the first time, the High Court has provided useful guidance on how the existing case law on the computer program patentability exclusion applies in the context of AI.

In this case, it was acknowledged that a computer program is needed to train ANNs but, crucially, the judge held that the claim did not claim that computer program; rather it was a "subsidiary part of the claim". Equally, it was held that the ANN itself was not a computer program. This allowed the computer program patentability exclusion to be circumvented despite the involvement of a computer program during the training stage. This is an important finding that suggests that the judge is mindful of preventing a lacuna of patent protection around ANNs (a key policy reason for the exclusion is to avoid duplicate protection for computer programs, which are generally protected by copyright).

The potential impact of this decision cannot be overstated, particularly as it has caused the UKIPO to review its examination guidance on patent applications relating to AI inventions. Whether the decision will be appealed remains to be seen. However, on the face of it, this decision positions the UK as an attractive jurisdiction for AI development where patent protection can be obtained, chiming with the UK government's ambitions for the UK to be at the forefront of AI development and its pro-AI innovation approach. The decision is also timely given the government's recent AI summit, significant investment in UK AI development, and the progress of UK and international initiatives to regulate AI.

Share

* This article is current as of the date of its publication and does not necessarily reflect the present state of the law or relevant regulation.

Connect with one of our experts

Interested in hearing more from Osborne Clarke?