A further decision from the European Patent Office’s Boards of Appeal illustrates the challenges faced by patent applicants when filing for patents for artificial intelligence (AI) inventions but also provides a number of suggestions on the level of detail that the patent office requires when disclosing the invention.
The decision T1191/19 (Neuroplasticity/Institut Guttman) from May 2022 relates to European Patent Application No. EP 2 351 523 that uses an AI approach to solve a problem in a medical setting. The application was refused based on a finding of insufficient disclosure in the description and a lack of inventive step over the prior art. The invention related to a method for selecting between different neuroplasticity interventions for patients based on a database of patients and the outcomes of the different interventions for the patients. The selection was performed using an AI and the patients were classified based on a ‘meta learning scheme’. The closest prior art was a scientific publication highlighting AI using the meta-learning scheme. Ironically, this prior art had been cited by the applicant itself to argue that the application was not unclear.
The Board of Appeal found that the application did not disclose how the meta-learning scheme was applied to the problem in a manner sufficiently clear and complete for it to be carried out by the person skilled in the art. The Board highlighted that the application did not disclose any example set of training data and validation data, which the meta-learning scheme set out in the description of the patent application required as input.
The Board highlighted that the description did not even disclose the minimum number of patients from which training data should be compiled to be able to give a meaningful prediction and the set of relevant parameters. The so-called “Heuristic Bases” A and B for training Classifiers A and B (“a respective set of heuristic or deterministic rules different from that of the other classifier” in step a1 of claim 1) and the Meta Heuristic for training the Meta Classifier(“[meta-classification based on] at least heuristic or deterministic rules” for the solution of the problem at hand are likewise not disclosed, nor is the structure of the artificial neural networks used as classifiers, their topology, activation functions, end conditions or learning mechanism (see also T 161/18, point 2 of the Reasons).
In the assessment of the inventive step (or obviousness of the invention), the applicant faced the issue that the meta-learning scheme was disclosed in the closest prior art. The Board saw nothing in the application beyond a mere reiteration at an abstract level of the scheme disclosed in the closest prior art. There was apparently no non-obvious detail written in the patent descritpion to differentiate idea from the prior art meta-learning scheme and applying the idea to the problem of selecting a neuroplasticity intervention.
Inventive Step and Sufficiency of Disclosure
The case was similar to an older case T0161/18, which we commented here. Both cases were found to be unpatentable for lack of sufficient disclosure in the application as a whole and lack of inventive step of the claims. These issues are caused by the same problem: there was simply not enough detail in the application to enable the skilled person to carry out the invention. The AI that is needed to carry out the invention was disclosed without enough detail that the skilled person could not carry out the invention. The Board states clearly, “at the level of abstraction of the application, the available disclosure is more like an invitation to a research programme.”
The lack of inventive step was caused by the same problem. The same absent disclosure needed to meet the sufficiency requirements could also have been used in the claims to describe how the AI is adapted to the problem at hand to solve the problem, such that the purported advantages are achieved across the whole of the claim scope.
Learnings
This decision serves as a reminder that the patent examiner expects applications for AI inventions to provide sufficient detail in their applications to enable the skilled person to produce a trained AI to put the invention into practice.
It is true that the Board indicated that a full set of training data should have been provided in the application to meet this requirement. In practice, however, it seems that a patent examiner will acknowledge that these requirements can be met with more limited disclosure. For example, experience shows that it is sufficient to describe how to collect a set of training data and, if necessary, validation data together with examples of such data. Furthermore, details on how to train the AI system, based on the training data, should suffice. These details should include an outline of the training method used, and the software in which the method was implemented, as well as any additional parameters that were selected. In Europe, new material cannot be added to a patent application after it has been filed as a European or PCT application. It is therefore often better to disclose as much as possible in the initial application as filed, erring on the side of “over-disclosure” rather than “under-disclosure”, even if the selection of the data and the software used for training the AI seems to be simple to the inventor.
Please feel free to get in touch if you have any questions. Our success rate in patent grants for AI inventions is high, and we’d be delighted to support you.
