“Improving the Chances of AI Patent Applications at the EPO: Insights from Haseltine Lake Kempner Partner”

Improving AI Patent Application Success at the EPO

In the rapidly evolving field of artificial intelligence (AI), the European Patent Office (EPO) has been proactive in anticipating a surge of AI-related patent applications. However, the process of preparing AI patent applications that stand the best chance of success in Europe requires careful consideration. In this article, Caroline Day, a partner at Haseltine Lake Kempner, sheds light on some of these considerations and how a well-drafted application can significantly improve the chances of patent approval.

AI as Computer Implemented Inventions

The EPO’s approach to computer implemented inventions (CIIs) has evolved over the years. While computer programs ‘as such’ are excluded from patentability by the European Patent Convention, CIIs that possess a ‘technical character’ are not. The EPO uses a ‘two hurdle’ approach to assess CIIs. First, it checks if a claim contains any technical features. Second, it evaluates the inventive step, allowing only technical features to contribute to the argument that the claim is not obvious. This two hurdle approach is also applied to AI inventions, but with additional considerations.

AI as Mathematical Methods

Many AI implementations can be characterised as mathematical models or methods. However, the European Patent Convention states that mathematical methods ‘as such’ are not patentable. To be patentable, an AI invention must ‘leave the abstract realm’. This could be achieved by applying the invention to solve a technical problem in a specific field of technology or by directing the invention to a specific technical implementation of AI. The EPO refers to these as the ‘two dimensions’ of technical character.

When preparing a patent application, it is crucial to consider these dimensions. For instance, the invention may be tied to particular hardware or use of specific hardware. A recent case (T2330/13) saw a method claim, initially objected to by the EPO as embodying a mathematical method, saved by being directed towards efficient parallel processing.

Another dimension to consider is whether the invention relates to a particular field of technology. For example, an invention may cover a conceptual change to the structure of an AI method that provides an improved output. However, the EPO may not consider this to have technical character unless it is associated with a specific technical application.

AI and Experimental Data

Experimental data has long played a role in EPO proceedings, particularly in the chemistry field. Such data can be used to demonstrate a plausible technical effect across a broad claim. Without this plausibility, an application may be restricted to the area of examples given. Experimental data can also defend a claim against an allegation of a lack of inventive step.

These practices, well established in relation to chemical patent applications, could also aid the patentability of AI inventions. Therefore, they are worth considering at the drafting stage of an AI patent application.

In conclusion, a well-drafted patent application that considers the specific requirements and considerations of AI inventions can significantly improve the chances of patent approval at the EPO. As AI continues to evolve and permeate various sectors, understanding and navigating the patent application process will be crucial for innovators in the field.

Caroline Day is a partner based in the Bristol offices of Haseltine Lake Kempner.