This article gives an important update on patent eligibility in Europe for computing-related inventions, including the latest announcements for Artificial Intelligence (AI) and Machine Learning (ML). We contrast the EPO position with the USA, and offer some practical tips to consider when drafting your next patent application in this field.
What’s happening in Europe on AI and ML inventions?
The European Patent Office (EPO) issued updated Guidelines for Examination, effective 1 November 2018, which for the first time discuss Artificial Intelligence (AI) and Machine Learning (ML). These new technologies fall within the existing long‑established framework and are treated under the same principles as other computing-related inventions. The overall legal framework on patent eligibility at the EPO has been established for over 20 years. The EPO’s position is relatively clear, giving confidence that applications will be treated fairly and with predictable outcomes.
How does Europe compare with USA?
The overall position on patent eligibility at the EPO for computing-related inventions is similar to the USA under Alice. An invention which is likely to receive a S101 rejection at the USPTO is also likely to be challenged for patent eligibility at the EPO. However, there are some important differences in the detailed analysis as conducted by each Office.The US system positively defines patent-eligible subject-matter (i.e. ‘process, machine, manufacture, or composition of matter, or any new and useful improvement thereof’), whereas the European Patent Convention (EPC) instead only states that certain categories are excluded, as listed in Article 52(2) EPC. The excluded categories are set out in the EPC, i.e. explicitly in the statute itself, rather than being developed in the Courts as in the US system. This fundamental difference underlies the two different approaches.
The US system has a single test for patent eligibility under S101, which is considered in isolation. Having met the S101 requirements, the analysis then proceeds to S102 for novelty and S103 for obviousness, without looking back. By contrast, the EPO system keeps patent eligibility in mind throughout examination, especially during the analysis of inventive step under Article 56 EPC.
At the EPO there is a low first hurdle as to whether the claimed subject-matter is completely excluded, which is easily overcome by using an appropriate claim formulation (e.g. by reciting A computer‑implemented method). However, for the detailed analysis, the claims and the description need to work together, in ways that might be unfamiliar to those preparing applications under US practice.
For many inventions (e.g. mechanical devices) the position on patent eligibility is clear from the outset and no further investigation is needed. However, when eligibility looks to be an issue, the EPO use a refined problem-solution analysis with steps (i), (ii) and (iii).
Step (i) - the low first hurdle
The first step at the EPO is to look at claim 1 as a whole and determine whether the subject matter relates solely to one of the excluded categories:
i) does the claim relate solely to excluded matter as in Art. 52(2) EPC?
For example, “A method of trading between aseller and a buyer by agreeing prices (a), (b) and (c)” relates only to a business method, and “A method of calculating a Fast Fourier Transform by mathematical steps (d), (e) and (f)”concerns a mathematical method per se. Business methods and mathematical methods are both specifically excluded under Article 52(2) EPC, and therefore both of these example claim formulations would be rejected outright by the EPO. However, this first hurdle is easily overcome by including at least one technical feature in the claim, i.e. by reciting “A computer-implemented method of trading between a seller and a buyer …”, or by mentioning technical features in the body of the claim, (e.g. ‘(a) calculating by a processor a first offer price…’. It might seem here that all we need is to say ‘using a computer’, but this is only the easy first step.
Typically, claims prepared under US practice will already include some technical features so as not to be dismissed as merely reciting an ‘abstract idea’ under step 1 of Alice, and hence also meet step (i) of the EPO test.
Steps (ii) & (iii) - the difficult second hurdle
The second phase at the EPO is much harder. Here the Examiner will assess the invention for inventive step under Article 56 EPC using the ‘problem-solution approach’, which is
ii) select the closest prior art document(‘D1’)
iii) identify the differences of claim 1 over that closest document D1, then:
(a) if there are no differences then the invention is rejected for lack of novelty
(b) if the differences relate only to the excluded categories then there is no technical effect and the invention is rejected for lack of inventive step
(c) if the differences involve a mixture of technical and non-technical features, then the non‑technical features become part of the ‘objective technical problem’ given to the skilled person.
The EPO rely on ‘technical effect’ as an essential part of their analysis of inventive step. At the EPO, an invention only has a technical effect when it makes a useful contribution in a technical field - in other words where the features of the claim result in a real-world outcome. In general terms this is similar to ‘something more’ as in 2(b) of Alice. However, that ‘something more’ needs to be understood from the specification itself and should be explicitly mentioned in the description.
Can you give more detail on the EPO analysis?
The Examiner will find the closest prior art document (typically referred to a document ‘D1’), which provides the most promising starting point for the skilled person to arrive at the claimed invention. When drafting the application it is impossible to know what that closest document will be, but it is possible to consider what it might be based on the expected background art, and to plan accordingly.
The Examiner will identify the distinguishing features of the claimed invention over document D1. In other words, the Examiner will isolate the features of claim 1 which are not disclosed in D1. The Examiner then considers the technical effect of those distinguishing features, i.e. asking why do those features achieve a useful result over and above the teachings of D1? At this point the Examiner will look in the description to understand the technical effect of those features. Therefore, the description needs to be written to support the technical effect of the claimed features.
Why do the description and claims have to work together?
Patent applications prepared under US practice are very good at describing the features of the invention in terms of what they do (e.g. ‘the processor outputs the encoded signal’). However, those applications can often lack any explanation of why those features have been provided (e.g. ‘so that only a specific client device can decrypt the signal, giving a more secure system’). This leads to difficulties, because the EPO need to understand the ‘why’ in order to know the technical effect of those features. Sometimes the ‘what’ also explains the ‘why’, but it is important to consider both aspects in the detailed description.
This often needs some creative thought, because the invention will have been conceived from a business point of view and explained in that way to the patent attorney. Further investigation is often needed in order to understand why those features actually lead to technical improvements, and then explain that in the detailed description.
A US patent application usually indicates the overall aim of the invention, e.g. to reduce cost, improve security, etc. However, the EPO also need to understand the effect of each individual feature which has been described, and why those features together contribute to the overall aim. Notably, the dependent claims are intended to provide fallbacks in the event that claim 1 is not novel. In which case, it is vital to also indicate the technical effect of those additional features in the dependent claims, because they may then be the only difference over the cited document D1.
What happens with claims having a mixture of technical and non-technical features?
The EPO analysis of inventive step will strike out from claim 1 all of the non-technical features (i.e. those relating to the excluded categories such as business methods and which don’t contribute to the technical effect). The question then is whether the remaining technical features are obvious to the notional skilled person. However, the skilled person is deemed to be given the non-technical features as part of their starting point. In other words, a non-technical client has written down their requirements and given that to the engineer to implement. If the technical implementation is straight forward then the invention must be obvious. This gives a ‘double strength’ effect – the non-technical features are themselves new, but are now being given to the notional skilled person and hence can’t contribute to non-obviousness. In the words of the Guidelines:
“Whether or not the new business concept is innovative is irrelevant for the assessment of inventive step, which has to be based on the features of its technical implementation”.
Having reached this point in the analysis, many of these mixed-type inventions are then rejected by the EPO on the basis that the distinguishing technical features are obvious, i.e. from other documents or routine practice in the art.
How does that apply to AI and ML inventions?
Artificial intelligence and machine learning are, at heart, mathematical methods – which is another of the excluded categories under the EPC. By themselves, AI and ML do not provide any technical contribution in themselves, in the same way that finding a square root is a purely mathematical function. Again, it is the why which is vitally important under EPO practice. The AI or ML has to serve a specific technical purpose, by being applied to a specific technical field or adapted to a specific technical implementation. For example, classifying data records or even ‘telecommunications data records’ without any indication of the technical use of that classification does not have a specific technical effect. Instead, the claims need to specify a real-world outcome, such as using aneural network to identify irregular heartbeats in a heart monitoring apparatus.
Can you give a simple checklist?
The new EPO guidelines have been welcomed by patent attorneys in Europe and are working well in examination with the EPO. We’d be happy to give further advice on any particular invention or application, including the best claim strategy for a positive outcome and strong protection.
The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.