Recent advances in artificial intelligence (AI) have resulted in a substantial increase in the use of AI in many fields of technology.
These new systems may include potentially patentable subject matter (inventions) and many of the national patent offices have been considering how these new AI-based technologies fit into existing patent rules.
The UKIPO published a guide for the examination of UK patent applications relating to AI in September 2022. This guide was updated in May 2024 and, most recently, in January 2025. The UKIPO has also published some example scenarios applying the guidelines to AI inventions:
The guidelines consider two key aspects of the Patents Act 1977 and how they apply to AI inventions. Specifically, these are excluded subject matter and sufficiency.
Excluded Subject Matter
Excluded subject matter refers to subject matter that is ineligible for patent protection. The most relevant excluded subject matter categories for AI inventions are “a program for a computer” and “a mathematical method”. If the AI invention makes a technical contribution, it will not fall under the exclusions.
The UKIPO guidelines explain that a program for a computer is not limited to programs for digital computers but to other kinds of computers, with analogue computers, artificial neural networks, hybrid computers and quantum computers being provided as examples.
Established case law considers whether an invention makes a technical contribution by applying the “Aerotel test”. The Aerotel test is a four-step test that aims to analyse the claim and consider whether its contribution falls solely within excluded subject matter or can be considered “technical”. For computer-implemented inventions UK Courts have developed “five signposts” which are provided as a guide to assess whether a technical contribution is present.
The UKIPO guidelines demonstrate the applicability of existing case law for the assessment of AI inventions, and the five signposts are as follows:
i) whether the claimed technical effect has a technical effect on a process which is carried on outside the computer;ii) whether the claimed technical effect operates at the level of the architecture of the computer; that is to say whether the effect is produced irrespective of the data being processed or the applications being run;
iii) whether the claimed technical effect results in the computer being made to operate in a new way;
iv) whether the program makes the computer a better computer in the sense of running more efficiently and effectively as a computer;
v) whether the perceived problem is overcome by the claimed invention as opposed to merely being circumvented.
It is noteworthy that the new guidelines have a section on AI models and data sets. The training of AI models may be referred to as machine learning. It is acknowledged in the guidelines that inventions involving the training of AI inventions are not excluded if they exhibit the required technical contribution.
Helpfully, the calibration of a technical device is presented as an analogy for training an AI. Specifically, it is stated that:
“Under the Aerotel approach, a computer-implemented method of calibration that makes a technical contribution to the art would be a patentable invention. By analogy, it follows that methods of training AI models or machine learning for achieving a specific technical purpose may also make a technical contribution.”
Sufficiency
Section 14(3) of the Patents Act 1977 requires that an invention is disclosed in a manner that is clear enough and complete enough for an invention to be performed by the person skilled in the art.
The new guidelines make it clear that existing methods for assessing sufficiency also apply to AI inventions.
Specific mention is made to the role of training data, and the extent to which it should be disclosed. A reference is made to an EPO Board of Appeal decision T0161/18 which is considered as consistent with the approach at the UKIPO.
In T0161/18 it was found that “the application does not disclose which input data are suitable for training the artificial neural network according to the invention, or at least one dataset suitable for solving the present technical problem”. It was concluded that the training of the artificial neural network could not be achieved by the person skilled in the art, and therefore the sufficiency requirement was not met.
Therefore, care should be taken to ensure that any patent applications relating to trained AI models clearly explain how the model was trained, including describing any specific data necessary for this purpose.
The Future
The latest update to the UKIPO guidelines was in response to a judgement of the Court of Appeal in relation to a patent application filed by Emotional Perception AI Ltd, which relates to an AI invention (specifically, artificial neural networks (ANNs)).
The next stage is for the case to be heard before the Supreme Court, meaning that the assessment of AI system patentability may shift in the near future.
Conclusions
It is clear that the assessment of AI inventions at the UKIPO is a developing field. However, it seems that the existing and well-established case law, particularly in relation to computer-implemented inventions, is applicable to AI inventions.