Sufficiency of AI inventions at the EPO

Under Art. 83 of the European Patent Convention, “[t]he European patent application shall disclose the invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art.”

 

This requirement has important practical consequences in the fast-developing field of AI.  If a patent application is filed before a new AI invention has been sufficiently developed, there is a risk that it fails for lack of sufficiency. Wait too long, and the invention might become public knowledge, meaning the opportunity to protect it is lost.

 

Case law in this field has mainly arisen in life sciences.  However, the question of sufficiency is increasingly relevant to AI inventions, particularly inventions concerning the use or training of ML models.

 

In T1191/19, the EPO Board of Appeal rejected an application concerning artificial neural networks on the grounds that it did not disclose sufficient information. The invention related to selecting between different neuroplasticity interventions for patients based on a database of patients and the outcomes of the different interventions, noting:

 

“The application does not disclose any example set of training data…and validation data…which the meta-learning scheme requires as input. The application does 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 Heuristic Bases A and B for training Classifiers A and B… and the Meta Heuristic for training the Meta Classifier … 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…At the level of abstraction of the application, the available disclosure is more like an invitation to a research programme.”

 

This 2022 decision has led to some uncertainty. Are these conclusions contained to the specific facts of T1191/19, or indicative of a general (and arguably high) standard of sufficiency for machine learning inventions? 

 

The EPO has brought welcome clarity in the March 2024 update to the Guidelines for Examination (G-II-3.3.1 Artificial intelligence and machine learning):

 

“The technical effect that a machine learning algorithm achieves may be readily apparent or established by explanations, mathematical proof, experimental data or the like. While mere allegations are not enough, comprehensive proof is not required either. If the technical effect is dependent on particular characteristics of the training dataset used, those characteristics that are required to reproduce the technical effect must be disclosed unless the skilled person can determine them without undue burden using common general knowledge. However, in general, there is no need to disclose the specific training dataset itself.”

 

This part of the Guidelines does not directly relate to sufficiency.  Rather, it relates to the requirements at the EPO for an invention to credibly achieve a technical effect under the assessment of inventive step.  Nevertheless, in practice the requirement to sufficiently disclose the invention and the requirement to demonstrate a credible technical effect across the scope of the claim for inventive step may be closely tied.  This link is made in another update to the Guidelines (F-III-3 Insufficient disclosure):

 

“Another example [of insufficient disclosure] can be found in the field of artificial intelligence if the mathematical methods and the training datasets are disclosed in insufficient detail to reproduce the technical effect over the whole range claimed. Such a lack of detail may result in a disclosure that is more like an invitation to a research programme.”

 

The update mirrors the language of T1191/19, but arguably qualifies the reasoning in the context of technical effect. If the technical effect achieved by an ML invention is readily apparent or explainable from the patent specification without experimental data or reference to a specific training set, that presumably increases the likelihood of reproducibility.

 

For further information please contact Tom Woodhouse, Patent Attorney.

 

This briefing is for general information purposes only and should not be used as a substitute for legal advice relating to your particular circumstances. We can discuss specific issues and facts on an individual basis. Please note that the law may have changed since the day this was first published in March 2024.