The EPO released its latest (2018) edition of the Guidelines for Examination on 1 October 2018. This release includes a number of updates in the Computer Implemented Invention (CII) field. We think there are five key points of which applicants should be aware.
Distributed computing environments
The EPO defines a ‘distributed computing environment’ as including the expected Cloud-based systems and client/server or base station/mobile device type environments. These environments will be relevant to a large number of technologies including Blockchain and augmented reality systems.
The new EPO guidelines confirm that it is possible to have multiple independent claims in the same claim category (e.g. apparatus, method) in a single application when claiming an invention that operates in a distributed computing environment. Each independent claim must be novel and inventive in its own right.
This is largely business as usual as it reflects the strategy that patent attorneys have been using for many years in order to seek protection for the various participants in a distributed system. However, the update expands upon this point by indicating that a claim to an invention that relies upon multiple distributed components to put the inventive concept into effect will be found unclear unless it the claim specifies which entity in the distributed environment performs each of the essential features of the invention.
This emphasises that careful drafting is required, particularly in the case of computer-implemented method claims, to ensure that this newly codified clarity requirement is met whilst simultaneously ensuring that the appropriate scope of protection is obtained. Selection of claim language will be particularly important in situations where it is not critical at which point within the distributed system a particular operation is performed. For example, in such cases it may be preferable to claim a generic ‘data processing device’ performing a particular step, rather than a more specific term such as ‘base station’, ‘cloud-based server’ or ‘user device’.
The EPO have provided a number of example inventions for mathematical methods that would be considered patentable. The essence of this section is that a pure mathematical method is not patentable, but a mathematical method that gains technical character by contributing towards producing a technical effect is patentable.
In practice technical character can be obtained by using a mathematical method to control a technical system or process, or by specifically designing the mathematical method to work effectively on a computer. Demonstrating that a particular mathematical method is more efficient than a known mathematical method will, however, not be enough to give the mathematical method technical character.
Given this latest guidance, it is advisable to describe (and ideally claim) an invention having a mathematical method in the context of the intended technical use of the method. Context is key – abstract descriptions of the invention that are divorced from a ‘real world’ use will cause the EPO to reject the application as non-inventive due to a lack of technical character.
AI and machine learning
The latest guidance makes clear that both AI and machine learning are being assessed by the EPO as specific sub-classes of mathematical method. The requirement of tying the mathematical method to a particular technical process or system to give technical character is thus being applied equally to AI and machine learning. Context is again key – link the AI or machine learning algorithm to use in a technical system to ensure that a technical effect is derivable from the application as filed.
Additionally, a notable update in this section of the Guidelines comes in the form of the final paragraph, where it is stated:
Where a classification method serves a technical purpose, the steps of generating the training set and training the classifier may also contribute to the technical character of the invention if they support achieving that technical purpose.
This relatively innocuous statement seemingly opens the door to claims covering the development and use of training datasets to train machine learning algorithms, which may have particular impact in fields such as automated drug discovery.
Simulations, computer-aided designs and modelling
The new Guidelines indicate that the EPO will treat these categories of invention as computer-implemented mathematical methods, such that the principle of tying the invention to a technical process discussed above is once again key.
The underlying message here is that applicants should look to describe and claim a simulation, computer-aided design or modelling-based invention in the context of the technical system or technical process.
It is worth noting that the new guidance makes clear that claiming the output of a simulation, design or model, such as a circuit manufactured according to an optimal design as identified by a simulation, is not necessary to obtain a granted patent.
Data retrieval, formats and structures
It has been confirmed in the new Guidelines that data structures have technical character when claimed as recorded on a computer-readable medium or as an electromagnetic carrier wave. However, an inventive step will only be acknowledged in cases where the particular data structure or format relates to functional considerations rather than cognitive considerations.
Data based on functional considerations is defined as relating to the control of the operation of a data processing device. Data based on cognitive considerations, on the other hand, is defined as relating to content and meaning that is relevant only to a human user.
As a result, it is advisable to focus on the aspects of an invention relating to control of a device when describing and claiming inventions relating to data structures. The wider human-relevant context in which the invention operates may still be included in the application, but this should not be relied upon to provide an inventive step.
If you wish to discuss patenting your computer implemented invention, please contact Sam Jones via email@example.com.