Tuesday, 5th February, 2019

European Patent Office releases guidelines on AI patentability

In November last year, the European Patent Office (EPO) released guidelines relating to the examination of patent applications for artificial intelligence (AI) and machine learning (ML) based inventions. These guidelines, G-II 3.3.1, provide clarification on how patent applications for inventions relating to AI and ML will be handled by the EPO.

The patentability of AI and ML based inventions has received interest of late, partially due to media hype and the growing expectations and concerns surrounding AI and ML technologies. In May last year, the EPO hosted an international conference to discuss the challenges of patenting AI and ML inventions. Last June, the heads of the patent offices of China, Japan, Korea, Europe, and the United States (the IP5) jointly noted that the impact of AI and ML on the patent system was one of their top strategic priorities.

The newly released EPO guidelines address the issue of the patentability of AI and ML based inventions. The guidelines provide that inventions relating to AI and ML are based on ‘computational models and algorithms for classification, clustering, regression and dimensionality reduction, such as neural networks, genetic algorithms, support vector machines, k-means, kernel regression and discriminant analysis.’

The guidelines outline that these inventions are ‘per se of an abstract mathematical nature’ and are therefore excluded from patentability under Article 52(2) as they lack technical character. Inventions of this nature have been excluded from patent eligibility for many years.

This aligns with the current similar approach for assessing computer-implemented inventions before the EPO.

Thus, the standard two-hurdle approach typically used for assessing such computer-implemented inventions also applies for assessing inventions relating to AI and ML. This approach requires (a) assessing whether the claimed subject matter has a technical character, and if so, (b) whether there is a non-obvious technical contribution over the prior art.

Therefore, AI and ML based inventions that are claimed in such a way that they can be considered to have a technical character and provide a technical solution to a technical problem would likely be looked upon positively by the EPO.

To further clarify this, the guidelines give the following favourable examples of patentable subject matter AI and ML applications: ‘a neural network in a heart-monitoring apparatus for the purpose of identifying irregular heartbeats’, which makes a technical contribution in the heart-monitoring sensor; and the ‘classification of digital images, videos, audio or speech signals based on low-level features (eg, edges or pixel attributes for images)’, which is a ‘technical application of classification algorithms’.

In contrast, the guidelines state that the following applications would likely not be patentable: ’classifying text documents solely in respect of their textual content’, or ’classifying abstract data records without any indication of a technical use being made of the resulting classification’, as these are not per se a technical purpose.

The new guidelines provide an assurance that the EPO is open to allowing inventions which relate to AI and ML, and also clarify concerns around the patentability of AI and ML based inventions, which ultimately should provide some certainty to inventors. As more and more AI and ML patent applications are filed, further examples of the boundaries of patentability will be created.