As the European commission announces increased investment and strategic coordination of AI innovation via the Horizon 2020 programme, the European Patent Office ("EPO") is now well used to handling patent applications for AI related technologies.
Our patent attorney teams continue to see significant numbers of patent applications directed to AI technologies, including to algorithms, networks, and training models in cutting-edge applications in various sectors like healthcare, medical devices, chemical fields, process automation, cyber security, space and defence, as well as AR, VR, computer vision, robotics, sensors, smart speakers, digital assistants, chat bots, and the like.
Stakeholders in artificial intelligence need to apply due consideration when selecting innovations for patenting as there are pitfalls for the unwary; and patent applicants should give particular thought as to how inventions should be framed having regard to the prior art. Embarking on patent filing in this field without a well thought out strategy inevitably turns out to be expensive and may also fail to provide the desired protection.
A first EPO hurdle requiring technicality tends to be relatively easy to overcome by claiming computer apparatus or a method reciting technical features.
However the real test for any AI related invention comes with the second hurdle of evaluating inventive step, namely in assessing whether the invention makes a non-obvious technical contribution to the art.
The EPO has for some years adopted a practice of disregarding non-technical features when it comes to assessing inventive step of a claim. Therefore aspects of an invention which look more like normal physical interactions between software and the computer that runs it, or mental acts, or mathematical methods, or administrative tasks, business logic, rules for playing games etc all face potential challenges, and particularly if claims are not if not framed sensitively.
A patent must also be clear and provide sufficient technical disclosure of what is required to implement the invention. Drafters should avoid marketing terms and buzzwords in the claims and take care to adequately describe specific implementations for functional features of the invention. Expressions such as "support vector machine", "reasoning engine", and "neural network" will be looked at in context to check the claim is properly framed around the solution of a technical problem, rather than merely reciting a mathematical model devoid of technical character.
Examples in current guidance indicate, for example, that (i) use of a neural network in a heart monitoring apparatus to detect irregular heart beats, and (ii) classification of digital images based on low level features (like edges or pixels) are both technical in nature. Whereas classifying text documents merely in respect of their textual context or classifying abstract data records ,without specifying a technical purpose is unlikely to be technical. Where technical purpose is served, the steps of generating the training model and training the classifiers may well contribute to the technical character and hence be claimable too.
The European Patent Office regularly allows correctly framed patent claims to AI related inventions that meet the novelty, inventive step and other substantive requirements. The EPO has trained numerous examiners to apply established principles based on case law from the various Boards of Appeals handling computer-related decisions. While there remain some grey areas the 2019 Guidelines for Examination go further than previous years in pulling together useful examples from the better-known case law decisions in the field of computer implemented inventions.
The Commission is increasing its annual investments in AI by 70% under the research and innovation programme Horizon 2020. It will reach EUR 1.5 billion for the period 2018-2020. It will: connect and strengthen AI research centres across Europe; support the development of an "AI-on-demand platform" that will provide access to relevant AI resources in the EU for all users; support the development of AI applications in key sectors.