The EPO and UKIPO approaches to AI and patentable subject matter
There are important ways in which the European and the UK patent offices differ when assessing the patentability of AI inventions, write Gemma Robin, Greg Ward and Ariana Sadr-Hashemi of Haseltine Lake Kempner in this co-published article
Inventions relating to Artificial Intelligence (AI), and more particularly to machine learning, have been raising questions for the patent community since they first started appearing as the subject of patent applications.
In particular, the thorny issue of the extent to which AI inventions may fall foul of the exclusions to patentable subject matter remains at the heart of many a headache for those working in the field.
The European Patent Office (EPO) has issued explicit guidance with respect to its approach to this issue, while the UK Intellectual Property Office (UKIPO) has announced that it will be releasing new guidance on examining the patentability of AI inventions under UK practice early this year.
EPO and UK approaches to excluded subject matter
The harmonisation of patent laws across contacting states to the European Patent Convention means that the EPO and UKIPO each start from the same list of categories of subject matter that are excluded from patent protection. For the purposes of inventions in the field of AI, the exclusions that most frequently appear are those relating to computer programs and to mathematical methods “as such”.
The wealth of European and UK patents granted for computer implemented inventions, as well as inventions underpinned by novel mathematical algorithms, is evidence that such inventions are eminently patentable. In all cases, the key is to ensure that the claims are not directed to an excluded category of subject matter “as such”, but rather to a solution to a technical problem, and that the computer program or mathematical method contribute to the technical character of the invention as claimed.
Both offices have well-established practices for considering applications relating to excluded categories of subject matter. In the following sections, we take a detailed look at the approaches of the two jurisdictions to excluded subject matter, and to AI inventions in particular.
The EPO approach
At the EPO, the approach to examining inventions for patentability is set out in the “Guidelines for Examination in the European Patent Office” (GfE), which embody case law from the EPO Boards of Appeal. As the number of AI related patent applications continues to grow, a dedicated section for AI inventions has been added to the GfE (G-II 3.3.1).
The EPO takes the approach that artificial intelligence and machine learning are based on computational models and algorithms that are in themselves of an abstract mathematical nature, and consequently the guidance for excluded subject matter provisions, and specifically for mathematical methods, should apply (GfE G-II 3.3.1).
Case law about what exactly is technical, and therefore patentable, is complex, but the EPO has clarified the two ways in which inventions can be technical - the two dimensions:
- The first dimension concerns a specific technical implementation (motivated by consideration of the internal functioning of a computer).
- The second dimension concerns a specific technical application (application to a field which the EPO deems technical).
When examining inventions that may contain a mixture of technical and non-technical features, the EPO’s preferred method is the Comvik approach (as set out in the EPO Board of Appeal decision T641/00).
Under the Comvik approach, only those features that contribute to the technical character of claimed subject matter are considered for the purpose of establishing an inventive step. Thus, the EPO will first determine which features are technical (according to one of the two dimensions). This will include features which, when taken in isolation, are non-technical, but in the context of the invention do help to produce a technical effect by interacting with explicitly technical features to serve a technical purpose. It is these identified technical features which may then contribute to establishing an inventive step.
Other features that fall exclusively within an excluded category, and do not in some way contribute to the solution of a technical problem, cannot contribute to demonstrating an inventive step for the claimed subject matter.
A consequence of the EPO approach is that objections to excluded subject matter will often arise in EPO prosecution in the context of inventive step, as opposed to an outright objection that the claimed subject matter falls within an excluded category.
The UK approach
In the UK, practice regarding excluded subject matter is based on case law from the English courts (the Aerotel/Macrossan test, as updated) and is set out in the “Manual of Patent Practice” (MoPP).
In addition to the MoPP, the UKIPO provides several “Enhanced Guidance” documents that cover particularly contentious areas of practice; the enhanced guidance document for AI inventions is currently in development and is expected to be made public soon.
As the number of AI-related patent applications continues to grow, a dedicated section for AI inventions has also been added to the MoPP (Section 1.39.3).
In contrast to the EPO approach of considering AI inventions in the context of guidance relating to mathematical methods, in the UK, while guidance relating to mathematical methods should be considered, “inventions involving AI are generally computer-implemented”. Consequently, such inventions should be assessed in accordance with the guidance on application of the Aerotel/Macrossan test to computer-implemented inventions.
The Aerotel/Macrossan test comprises four steps:
- properly construe the claim;
- identify the actual contribution;
- ask whether it falls solely within the excluded subject matter;
- check whether the actual or alleged contribution is technical in nature.
In practice, under the Aerotel/Macrossan test, if a claim is found to have a contribution that falls within a category of excluded subject matter, the application may not receive a search. That said, the UKIPO has indicated its intention to reduce the number of applications for which no search is performed.
Different approaches, different results?
The differing UK and EPO approaches to considering excluded subject matter are long established and the English courts have consistently concluded that, while different, they usually reach the same result in practice (for example, as confirmed by Birss J in paragraph 9 of Lenovo (Singapore) PTE Ltd v Comptroller General of Patents).
Anecdotally however, and specifically within the field of AI inventions, it seems that the day-to-day experience of attorneys prosecuting European and UK cases under examination can be quite different in the two jurisdictions.
The UKIPO has gone to considerable lengths to investigate outcomes of AI patent families prosecuted at the UKIPO and EPO. It has concluded that in terms of eventual outcome, there is minimal difference between the two. As this similarity in outcome doesn’t necessarily align with the personal experience of attorneys prosecuting the relevant cases at first instance examination, we have looked into a sample of prosecution histories for patent families in which comparison of UK and EPO practice is possible.
Perhaps unsurprisingly, given the restrictions against double patenting, the pool of published AI-related patent families in which both a GB and EP family member have entered examination is relatively small. The pool of suitable patent families in which examination is sufficiently advanced to draw useful conclusions is even smaller, making it difficult to provide definitive answers based on our review.
For the purpose of our research, we have considered published patent families from the last 10 years in the International Patent Class (IPC) G06N: “Computer Systems based on specific computational models”.
Of the patent families we identified fulfilling our requirements (GB and EP family members having entered examination), significant objections relating to excluded subject matter were raised on 70% of the GB cases. In many cases the objections were raised at the search stage, indicating major issues to be addressed at examination. At the EPO, meanwhile, excluded subject matter objections of comparable importance were raised on only 30% of the EP family members, primarily in the context of an inventive step objection; indicating that the novel technical features were entirely in an excluded field.
While the evidence base is limited, this disparity suggests that the anecdotal experience of attorneys prosecuting AI inventions at the EPO and UKIPO may be consistent with a larger trend. In particular, our analysis suggests that excluded subject matter objections for AI cases have in the past been raised more often, and more extensively, during UKIPO prosecution than during EPO prosecution. While the UKIPO analysis of prosecution outcomes suggests that this disparity is corrected in the end, our investigation into prosecution experiences for AI cases may provide some explanation for the impression that the UKIPO may have been, until now, a less favourable jurisdiction for AI subject matter.
The UKIPO is aiming to be a forum of choice for AI cases, and to this end it has undertaken wide ranging consultations and roundtable events. The office’s upcoming Enhanced Guidance, informed by the results of these public consultations and events, as well as close collaboration with the EPO, is one encouraging outcome of the UKIPO focus on this subject area. It suggests things may be looking up for patentees in the AI sector looking to protect their technology in the UK.
We look forward with interest to the publication of the Enhanced Guidance, and to considering its potential implications for the choice of where to pursue patent protection for AI inventions.
Gemma Robin is a Bristol-based partner in the Haseltine Lake Kempner AI Team. Greg Ward is a senior associate in the firm’s AI team and Ariana Sadr-Hashemi is a trainee patent attorney in the Tech Team; both are based in London.
Previous articles by Haseltine Lake Kempner authors in this series can be accessed here:
How to secure AI patents in Europe
Drafting AI patent applications for success at the EPO – eligibility and claim formulation
Drafting AI patent applications for success at the EPO – drafting the full specification
Technology trends – why patent your hidden AI?
Google and Samsung top the list of applicants for AI-related patents at the EPO