The so-called “Achilles heel of Artificial Intelligence” and the European Patent Office decision in Äquivalenter Aortendruck case


On 12th of May 2020, the EPO Technical Board of Appeal has issued decision T0161/18 upholding a rejection from the Examining Division for lack of inventive step (as stipulated in Article 56 EPC) and for lack of sufficient disclosure of training data (as stipulated in Article 83 EPC).

This decision has caught the attention of the public and the professionals as it related to a method involving artificial neural network.

The main claim of the application regards a method for determining cardiac output from an arterial blood pressure curve measured at the periphery, such curve being transformed into the central blood pressure curve by the aid of an artificial neural network whose weighting values are determined by learning and the cardiac output is calculated from the central blood pressure curve.

Among the reasons for rejection, we find reason 2, which entails lack of sufficient disclosure under Article 83 EPC. Here, the Board asserts that the application fails to disclose the input data to be used for training the neuronal network or at least a dataset enabling the technical problem to be solved.

As the Board states, this application brings only a vague mention as to the input data and its necessity to cover a broad spectrum of patients of different ages, sexes, constitutional types, health conditions and the like to avoid specialization of the network. Due to the lack of extensive information, skilled persons are not able to carry out the training of the network. In this regard, the Board concluded that the application as a whole does not meet the requirements of Article 83 EPC.

Why is this aspect so important?

Sufficient disclosure enabling the invention to be actually carried out is a quid pro quo for the award of an exclusive right to an invention that meets the patentability conditions thus fulfilling a public policy objective.

This decision can be regarded as a milestone in this area of law as it is the first Board of Appeal rejection decision for lack of disclosure in training data pertaining to an AI case.

Furthermore, another aspect that is worth mentioning is the fact that the Board raised this ground of insufficient disclosure of its own motion, an evident opposed background of review by the EPO Examining Divisions pertaining to compliance with Article 83 EPC.

Let’s take a view of the pharmaceutical and chemical sectors, as well as other sectors such as mechanical, physical or computer technology.

In the pharma sector, a claimed invention most often includes a broad family of species for which there is a desired effect, and since there is no logical reasoning linking the structure of these given species and the desired effect, the assessment regards whether it is plausible for the desired effect to be achieved across the entire family stipulated in the claim.

In other sectors, such as mechanical, telecom, etc., the Examining Divisions usually give slight attention to the compliance with Article 83 EPC, its issues occurring when they are raised by third parties in opposition proceedings or observations, and not by the motion of the Division.

The rationale behind this is the fact these third parties possess specific expertise, hence allowing them to spot insufficient disclosures that the Examiners might not detect. This leads to the conclusion that usually patents may be granted in spite of a total lack of relevant disclosure on a critical component. This situation favors speculative patenting.

As to the field of Artificial Intelligence, EPO alongside other IP5 Offices released a statement regarding the disclosure requirements in AI cases. These requirements are included in the Report from the IP5 expert round table on artificial intelligence Munich, 31st of October 2018, stating: “All IP5 Offices have strict disclosure requirements, including reproducibility and repeatability. However, the application of the requirement of sufficiency of disclosure allows for some flexibility.” This “strict disclosure requirements” is a strong departure from the scant attitude generally adopted by the Examination Division.

The future development of the BoA jurisprudence shall prove more guidance as to which disclosure can be considered to comply with the sufficiency requirement of Article 83. Furthermore, future Guidelines on this matter shall incorporate the Board’s reasoning and provide additional insights in order to help practitioners to meet these conditions, especially those involved in the field of training data. This aspect is highly important as practitioners intend to meet these requirements without providing public access to the data, due to confidentiality concerns.

In the past, a similar issue was raised in the case of inventions using software as the insertion of the source code of the software was sometimes used, however being considered sufficient to disclose the architecture of the software and the sequence of operations in such detail that a programmer could write the source code.

As to AI training data, a sufficiency condition can be considered to be met when the application discloses the actual methodologies for the selection of data sources and processing of data, specifically adapted in order to enable a skilled person to prepare this training data relevant to the objective.

However, there is a significant need for the EPO management to enhance the expertise of its Examiners in matters of AI technology because a proper assessment of the disclosure technology is critical taking into account the fact that AI technology has a rapid development and is constantly implemented in various technological fields.

Attorney at law Daniela Manolea