Patenting AI innovations in the UK

Why should I read this?

It is therefore unsurprising that there has been considerable discussion over the last 18 months on how the technology is to be regulated and deployed. Intellectual property concerns have also been a key part of the debate, particularly in relation to how various foundational models have been developed and trained using disparate data-sets. What has been less prominent in the AI debate so far is how AI innovations and the resultant intellectual property in AI can be adequately protected to provide market exclusivity for those making substantial investments in the technology. As previous technology booms have shown, providing sufficient market incentive for investment is often central to technology uptake and commercialisation.

There have been two recent UK cases surrounding patenting of AI innovations. In Thaler v Comptroller General of Patents Trade Marks and Designs, the UK Supreme Court (the “UKSC”) considered the requirement of human inventorship for patentability of AI inventions. In Emotional Perception AI Ltd v UKIPO the High Court grappled with the patentable subject matter exclusion for computer programs. In both cases, the judgments evaluated the historic application of specific provisions of the UK Patents Act 1977 (the “Act”) and how they may shape the approach to patenting AI innovations in the UK.

The requirement of “human” inventorship

The key issue before the UKSC in Thaler was whether the creator of an AI engine is entitled to a patent for inventions which have been autonomously devised by AI.

The appellant, Dr Thaler, had filed two patent applications designating the inventor as an AI machine called DABUS, on the basis that DABUS had independently identified and devised two “novel inventions”. Dr Thaler contended that the “inventor” for the purposes of the Act should be construed broadly to include a machine, notwithstanding the implicit presumption that inventors could only be natural persons. It was on this point that Dr Thaler based his arguments rather than claiming that the DABUS possessed any legal personality.

The scope and meaning of “inventor” in the Act:

The UKSC unanimously ruled that the meaning of “inventor” under the legislation was to be restricted to a “natural person”, and the grant of a patent in the UK was limited to “natural persons”. As DABUS was a machine, the concept of inventorship could not be extended to it, and as DABUS did not qualify as an “inventor”, then neither could its creations be classed as “inventions” for the purposes of the Act.

Dr Thaler argued that the inventions in question were the “fruits” of his DABUS machine and that just as “the farmer owns the cow and the calf”, Dr Thaler, as the owner of DABUS, possessed all rights in DABUS’s creations. The Court rejected this argument by reasoning that this did not obviate the central issue in contention, which was the need for the patent applications to designate an “inventor” within the meaning of the Act. DABUS, as a machine, did not meet the relevant requirements.

The UKSC did acknowledge that had Dr Thaler designated himself as the inventor on the application, and stated that he had used DABUS as a “highly sophisticated tool”, the outcome of the application "might well have been different”.

In dismissing the appeal, the UKSC referenced the opinion of Court of Appeal Justice, Lady Elisabeth Laing:

"Whether or not thinking machines were capable of devising inventions in 1977, it is clear to me that that Parliament did not have them in mind when enacting this scheme. If patents are to be granted in respect of inventions made by machines, the 1977 Act will have to be amended..."

The exclusion to patentability of “programs for a computer”

Another underlying issue which had been viewed as a potential bar to patentability of AI innovations is the subject matter exclusion for patenting “a program for a computer” as such. The scope of this exclusion in the context of AI inventions was considered by the High Court in Emotional Perception v UKIPO.

It has long been established that in order to overcome the exclusion and gain patentable status, inventors of computer programs have had to demonstrate, in line with the Aerotel case, that their invention provides a “technical contribution” which has a real world effect. This is a tangible effect or output other than one relating to the functioning or processing capability of a computer.

AI presents a unique question to the courts when applying the meaning of “technical contribution” and “computer program”, as its tangible real world effect represents a seismic change in technological capabilities and output to what was envisioned when the legislation was drafted. Crucially, the High Court in Emotional Perception took it upon itself to provide initial answers to these previously unexplored questions.

1. Can the training and operation of an artificial neural network (“ANN”) be classed as a computer program?

The Court held that an ANN operates at a layer above the software platform used to run computer programs.

The Court understood that the functioning of an ANN relies upon the replication of virtual nodal layers, which facilitates its analysis. The data analysis is self-learnt, within the parameters set, and in this case was vector-based, meaning that similarities in the distance between vectors could be mapped to identify similarities in the musical data. As stated at paragraph 50 of the judgment, “There was no program at that point because no person had given a set of instructions to the computer to do what it does – the ANN had trained itself”.

Consequently the ANN did not require human oversight to function and was not a computer program identified by the Court as “a set of instructions that makes a computer do a particular thing”, meaning the exclusion did not apply.

2. Does the invention make a technical contribution?

The Court also went on to consider the technical contribution of the invention to assess whether the invention could nonetheless be patentable, in the event that the ANN was classified as a computer program.

The Court held that the musical recommendation sent to users was a result of the invention facilitating the transfer of data analysis outside the computer system. This was to be regarded as an external real world effect and consequently a technical contribution in line with the previous caselaw, meaning it was not a computer program “as such”.

The invention in Emotional Perception

The invention submitted by Emotional Perception AI Limited claimed the method of training an Artificial Neural Network (“ANN”) through the use of tailored source files and natural language processing.

In this case the training files consisted of music, which would allow the ANN to map the similarities between music listened to by a user and music available to the user, in order to send tailored recommendations based on emotion and human perception of the songs rather than simply recommending songs in the same genre as existing websites do.

The implication of these decisions for AI innovation

The UKSC and High Court decisions are indicative of some of the challenges AI presents when considering the inventorship of outputs and their real world contributions. The ability of AI technology to produce output with limited (and sometimes no) human involvement means that the existing concepts in patent law (which are decades old) will be necessarily tested.

Emotional Perception AI Limited

The High Court’s finding of a “technical contribution” in Emotional Perception acknowledges the capability of AI training methods to produce a twofold internal and external effect. The recognition that such training methods are geared towards the production of user-specific outputs via the adaptation of internal data processing, takes them beyond the statutory exclusion and creates a precedent which will allow for the finding of a “technical contribution” to be more readily available.

This step is crucial in fostering greater development in the field, as deep learning systems such as ANNs rely heavily on specified and tailored data file training which set the parameters of data analysis and allow the ANN to deliver the inventor’s envisioned output. Developers will be encouraged by this more permissive approach.

Since the decision the UK Intellectual Property Office has updated its statutory guidance on examination of patent applications involving ANNs, requiring patent examiners not to object to inventions involving the training of an ANN under the “program for a computer” exclusion. It should be noted though that Emotional Perception is a first instance decision and it will interesting to see how these issues are treated at the appellate level.

Thaler

The UKSC’s judgment in Thaler confirms the requirement for a “natural person” to be designated as an inventor for the purposes of applications relating to AI output. The UKSC’s acknowledgement that had Dr Thaler designated himself as the inventor on the application the outcome may well have been different, demonstrates that the Court is not hostile to the patentability of AI inventions. Rather, it simply requires applications for such inventions to be filed in a manner which is consistent with the statutory requirements.

The UKSC’s ruling on the need for human inventorship was widely anticipated and is consistent with the approach taken in most other jurisdictions. The decision reinforces the position that under existing law, inventions which are entirely machine generated are not patentable without the designation of a human inventor. It is possible that changes in the law to allow for computer program inventors will be eventually introduced, though the UK government has ruled this out for now.

Of course, the current debates around AI patents must be understood in the context of the broader discussion around IP protection more generally. In this regard, it is likely that know-how/trade secrets along with copyright will continue to be important to protecting AI innovations. In addition, contractual considerations - such as restrictions around the field of use, confidentiality, and the need for adequate disclosure in technology licenses – will also be key in protecting an innovator’s commercial interests. A strategic and holistic approach to IP protection and commercialisation is therefore critical as the AI boom takes off.

How can Eversheds Sutherland help?

Whilst the decisions and updated UK Intellectual Property Office guidance bring a degree of clarity to the world of patentability and AI, it is still very much a developing area. Both decisions highlight the need for expert advice and consultation prior to filing and on IP strategy more broadly (see our previous articles here and here). Our experienced team would be happy to provide any assistance you may require in navigating this exciting and complex field.