This browser is not actively supported anymore. For the best passle experience, we strongly recommend you upgrade your browser.
| 2 minutes read

UKIPO issues guidance on the patentability of AI inventions

The UKIPO has issued long awaited guidance on the patentability of AI based inventions. The guidance appears to be fairly comprehensive, with 18 example scenarios of patentable, and excluded, AI based inventions.

Of particular interest is the section on "Core AI". Core AI is defined in the guidance as not being adapted for any specific application, but instead defining an advance in the field of AI itself (eg, an improved AI model, algorithm, or mathematical method). Where an AI invention is applicable to many different applications, such as image processing, speech processing, etc., it is preferred to obtain patent protection for the invention without reference to a specific application (e.g. protect the AI model itself and not its practical application). This is usually a difficult task in the UK, as there is little UK case law on positive examples of patentable Core AI.

The new guidance provides three scenarios (16, 17 and 18) as examples of allowable Core AI. Each case relies on the third signpost of the "AT&T Signposts" as set out in AT&T/Cvon, which asks, does the claimed technical effect result in the computer being made to operate in a new way? If the answer is in the affirmative, the invention is likely more than a computer program as such.  As an aside, I've never been able to find a positive example of the third signpost in the case law, and the guidance echoes this by stating that "[u]nfortunately, there are few (if any) explicit examples of the positive application of signpost (iii) in decided case law". Therefore, the scenarios provided in the guidance are most welcome.    

Each scenario is relatively short and so I leave it to the reader to review, but note that the UKIPO will acknowledged a technical contribution where a functional unit of the computer (e.g. GPU or special-purpose accelerator, etc.) is made to work in a new way, or there is a new physical combination of hardware within the computer and that the AI running on the computer brings about a change to the technical operation of the computer relative to a conventional computer, not just a change to the way a program or algorithm works. 

What does this mean in practice? It appears that the UKIPO's new guidance on AI inventions corresponds to the EPO's guidance on mathematical methods. As such, arguments that have worked in Europe for Core AI based inventions should equally work in the UK. In particular, a non-application specific AI invention that interacts with the hardware of the computer in a new way will arguably not be excluded from patentability for being a computer program as such in the UK. Therefore, when drafting specifications to AI inventions, it will be beneficial to describe how the AI interacts with the hardware. For example, can the AI be run in parallel on, for example a CPU and a GPU, with certain portions of the AI preferentially executing on the CPU or the GPU? If so, it is worth describing this in detail as it may be that by doing so, it can be argued that the invention causes functional units of the computer to work in a new way.    

In summary, the new guidance, and in particular the scenarios, are most welcome and should help applicants and patent attorneys alike in navigating the complex world of excluded subject matter as it pertains to computer programs as such.

We consider that a core AI invention may make a technical contribution if its instructions define: a functional unit of a computer (e.g. GPU or special-purpose accelerator, etc.) being made to work in a new way, or a new physical combination of hardware within the computer (see paragraph 59 above), provided that the instructions produce a technical effect within the computer that does not fall solely within the excluded subject matter of section 1(2)


artificial intelligence, patents