Patent Power Half Hour is Kilburn & Strode’s information-packed, supercharged webinar series. We dive into typical stumbling blocks at the EPO and share practical tips to avoid them. All delivered in 30 minutes or less!
Here is a summary of the key take-aways from the webinar that ran on 25 May 2021, covering patenting AI/ML and mathematics in Europe.
In the last session we discussed how important it is, when patenting software-related inventions in Europe, to have a narrative around the technical character of your invention and to capture this in your application as filed.
This is the same with inventions in the AI/Machine Learning and mathematical space.
For inventions in this area, only the features which contribute to technical character are to be considered when assessing inventive step. This is where having the technical character narrative in your application proves invaluable. Make sure you have explicit basis in the application as filed linking a feature with a technical effect. Without this information it can be very hard to amend your claim to establish the technical character of your claim.
The EPO Guidelines make it clear that AI/ML is currently treated by the EPO as a mathematical method and that mathematical techniques such as AI/ML can contribute to technical character.
The two ‘safe harbours’
Two alternative lines of argument have been built up in EPO case law which applicants can rely on when ensuring a mathematical method, for example an AI method, contributes to the technical character of an invention. We like to refer to these two lines of argument as the two “safe harbours” for mathematical and AI/ML inventions at the EPO.
The first safe harbour relates to the technical application of the mathematical method. In summary, the EPO considers that an AI/ML or mathematical method may contribute to the technical character of an invention by virtue of its application to a specific field of technology. To ensure your claimed AI/ML or mathematical method benefits from this safe harbour, you need to be able to argue that the method, in the context of the invention, serves a technical purpose. This technical purpose needs to be sufficiently specific. Examples from the case law and Guidelines include digital audio, image or video enhancement or analysis, speech recognition, encoding data for efficient transmission, optimising load distribution in a computer network, and providing a medical diagnosis. Further, the claim needs to be ‘functionally limited’ to the technical purpose. In other words, it isn’t sufficient to simply define a technical purpose in the preamble of your claim. Instead, the mathematical method should be sufficiently linked to the technical purpose, for example by ensuring the claim defines the processing of data that is relevant to the technical purpose, and that the resulting output clearly contributes to the technical purpose.
The second safe harbour relates to the technical implementation of the mathematical method. The key consideration here is whether the method or algorithm design is “motivated by the technical considerations of the internal functioning of the computer”. For applicants prosecuting AI/ML at the EPO, this is an important sentence to remember. In other words, if a step of a method has been designed in such a way as to be particularly suitable for being performed on a computer, it arguably has a technical character. For example, if a method has been intentionally designed in order to reduce the amount of data being stored or transmitted, in order to make use of the advantages of particular digital circuits such as barrel shifters or programs such as pixel shaders, or to capitalise on efficient task sharing between CPUs and GPUs, the method may well benefit from the protection afforded by the second safe harbour.
Future- (and present-) proofing your application
So, you can patent AI/ML technology applied to a technical field or purpose or when it has been designed with the internal functioning of a computer in mind (some examiners err on the side of caution and require this to be special purpose hardware, while others acknowledge that this is not necessarily so). But what if your invention is a fundamental improvement to an AI/ML technology, such as a class of algorithms, architectures or training approaches?
Fundamental AI/ML innovation is considered to relate to mathematical methods as such, unless it can be limited to clearly show hardware considerations or a technical field of application. The former may simply not be the case and the latter is unlikely to do justice to the contribution to the field. So, what should you do to obtain strong patent protection for AI/ML inventions?
Our recommendation is to make specifications as comprehensive as possible or at least as comprehensive as justified by the commercial context. Technical applications? Yes, include as many as you can in as much detail as is practical. Algorithm design motivated by computer considerations? Great, these are always worth looking for. Faced with a fundamental contribution to AI/ML? Congratulations! Even though unlikely to be patentable outside the safe harbours for now, this may change as technological reality drives changes in the appeal case law (as has recently happened for database management technology).
So, to future proof applications containing fundamental AI/ML advances, it may be worthwhile drafting claims to any fundamental AI/ML innovation per se while not forgetting the additional details needed to benefit from the safe harbours already open at the EPO. This way, you can future-proof and present-proof your AI/ML application.
Stronger, faster and cheaper patents in Europe
What action can you take now? You could:
Review and update your guidance to US outside counsel regarding information to include during drafting patent applications involving AI/ML or mathematical methods
Review and update invention capture forms to capture details on the technical application and the technical implementation of your AI/ML or mathematical inventions
Consider co-drafting with a European attorney to ensure appropriate language is included in the draft to provide the necessary narrative around technical character
Consider pre-filing reviews and amendments before filing in Europe
Consider a go/no-go assessment to allow you to concentrate your resources on those cases that are more likely to succeed in Europe.
If you'd like our help with these suggestions or have any questions on this topic, please get in touch.
To suggest topics for future Patent Power Half Hour webinars, contact us via Insights@kilburnstrode.com.
Sign up to the next Patent Power Half Hour ►