On March 1, 2024, updated Guidelines for Examination (“Guidelines”) in the European Patent Office (“EPO”) went into effect. The updated Guidelines included substantive changes with broad implications for the patentability of artificial intelligence (“AI”) innovations. These changes will impact prosecution of all patent applications at the EPO going forward. More importantly, however, than the immediate impact of the Guidelines, the updated Guidelines demonstrate further hostility toward the patentability of AI technology in what is already one of the world’s toughest jurisdictions for that crucial technology.
One of the key changes in the updated Guidelines is the potential requirement to include specific details of the training data for a claimed AI model in a patent application. This requirement is tied to both insuring sufficiency of disclosure and as proof of technical effect. See Guidelines, F-III, 3 and G-II, 3.3.1.
Specifically, the Guidelines at F-III, 3 state that a patent disclosure is insufficient when “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 guidelines at G-II, 3.3.1 state that “the technical effect that a machine learning algorithm achieves may be readily apparent or established by explanations, mathematical proof, experimental data or the like. ... 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.”
While the Guidelines are clear that, at present, the full disclosure of the underlying training dataset is not generally required, this language indicates that there are circumstances in which such disclosure of the training dataset is required. Further, the circumstances under which full disclosure of the training dataset is required are not clearly identified. This ambiguity leaves applicants unsure as to what must be included in a patent application and also unsure as to whether the disclosure requirements of the EPO are too high to even justify filing at the EPO. Given the retroactive nature of the Guidelines in applying to previously filed applications and the extended duration of prosecution, the continued hostility of the EPO toward AI technology negatively impacts the expected return on investment for future AI patent applications at the EPO.
The true impact of the updated Guidelines will only become more apparent as the EPO provides examiner training for these new guidelines and through further decisions by the EPO Boards of Appeal. While many hope that the EPO will find innovative solutions to protect AI, the current guidelines underscore the necessity of a robust and up-to-date strategy for pursuing patent protection of AI technologies.
