Patentability of AI inventions in Europe: Life Sciences, Chemical and Biotechnology inventions

Whilst there may still be hurdles to overcome, and despite some misconceptions, AI inventions can be patented in Europe and have found particular application in the Life Sciences and Biotechnology sectors. Here, we discuss some of the challenges that might be faced when patenting these inventions before the EPO, and some ways these can be overcome.

AI inventions at the EPO Generally

The EPO employs what is commonly referred to as a “two hurdle test” when assessing patentability of such inventions.

Firstly, Article 52(2) of European Patent Convention (the EPC) defines the following categories as “non-inventions”:

1. discoveries, scientific theories and mathematical methods, 

2. aesthetic creations, 

3. schemes, rules and methods for performing mental acts, playing games or doing business, and programs for computers, 

4. the presentation of information. 

Any of this subject-matter, when taken “as such” (alone), is considered non-technical. 

Here, computer implemented inventions such as AI and machine learning are not explicitly stated in Article 52(2). But, in their Guidelines for Examination, the EPO defines AI and machine learning as being based on:

“…computational models and algorithms for classification, clustering, regression and dimensionality reduction, such as neural networks, genetic algorithms, support vector machines, k-means, kernel regression and discriminant analysis.”

But, when applied to AI and machine learning applications, the EPO typically defines this subject matter to be of an abstract mathematical nature. Therefore, the above noted “a)” applies: the mathematical method exclusion generally applies to AI and machine learning inventions when taken ‘as such’. Although if claims are drafted broadly, or in a certain way, these inventions can also be considered to cover a mental act carried out by a human.

So, to be considered patentable at the EPO, there must be something more to the invention than just being one of the listed “non-inventions” as such. It is established in European case law that if the invention has at least one “technical” component, then it does not fall within the exclusion. Claims with a technical component will not be objected to under Article 52(2).

Regarding what this means: when the claimed invention is based on a mathematical method, consideration will be made whether the mathematical method contributes to the technical character of the invention, i.e. does it contribute to producing a technical effect that serves a technical purpose? The two main criteria for a mathematical method to contribute to the technical character of an invention are the application to a field of technology and/or by being adapted to a specific technical implementation. There is a range of case law and examples related to these. 

This is the “first hurdle” of the “two hurdle approach” eligibility test at the EPO. The “second hurdle” relates to all of the usual criteria of patentability before the EPO, for example, novelty, inventive step and sufficiency. 

Particularly for assessment of inventive step, if there is a mixture of technical and non-technical claim elements; at the EPO this is assessed using the COMVIK approach, as we have discussed previously.

A further hurdle for patenting these inventions at the EPO that is particularly worth noting is ‘sufficiency of disclosure’. Article 83 EPC states that “The European patent application shall disclose the invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art.” This applies to assessment of all patent applications before the EPO, but can be particularly relevant to Life Sciences, Chemical and Biotechnology inventions when utilising AI, as discussed below in some relevant examples.

AI in Life Sciences, Chemical and Biotechnology inventions

As in all areas of technology, there has been a surge of AI and machine learning inventions being developed in these sectors. For example, AI is being utilised in diagnostic, genomics and particularly in drug discovery pipelines. Naturally, patent applications are an important part of this innovation. 

All the above discussed criteria apply to assessing AI-related inventions in these spaces, including the need for claims to have a technical character. In the EPO’s Guidelines for examination list of ‘examples of technical contributions of a mathematical method’, there are a number of technical contributions related to these fields: 

• Controlling a specific technical system or process, e.g. an X-ray apparatus;

• Digital audio, image or video enhancement or analysis, e.g. de-noising, detecting persons in a digital image, estimating the quality of a transmitted digital audio signal; 

• Providing a genotype estimate based on an analysis of DNA samples, as well as providing a confidence interval for this estimate so as to quantify its reliability; and

• Providing a medical diagnosis by an automated system processing physiological measurements.

Therefore, there is scope for a technical characteristic to be applied to these inventions, overcoming the ‘first hurdle’. 

Looking at some considerations for some specific examples in these sectors:

AI and drug/medicinal product discovery

Patents may offer powerful protection in the field of drug/medicinal product discovery, but there is much to consider here. Consideration is needed if patent protection should be sought for the AI methods, tools or processes used to generate the candidates, or just the resultant candidates themselves. 

In our experience, these methods can be patented i) if they can be directed to a workflow or therapeutic design workflow via the implementation of technical features, ii) if they can be specified as computer-implemented methods, and iii) if the claim language ensures that there is not an attempt to cover a mental act which could also be carried out by a human (also exempt from patentability under a different point of Article 52(2), noted above). 

For AI tools used in drug/medicinal product discovery, consideration should be given to if you want to publicly disclose the method when perhaps it could be kept as a trade secret. What code is involved, will this be constantly evolving, do you want to disclose this or an older version of it for competitors to work off, and is the code or a version of it necessary for a patent application considering sufficiency? Finally, would ‘method’-type claims actually be enforceable in this field? 

It is very likely that those working in this field will want to patent the products (for example drugs or vaccines) that result from such methods. In the pharmaceutical sector, patents (along with other intellectual property (IP) rights such as regulatory data exclusivity) are an especially key part of encouraging and protecting innovation. 

A product resulting from an AI method, tool or process is not inherently patentable or un-patentable; it will be assessed using all the usual requirements for patentability of a standalone product by the EPO. 

As with traditional methods/processes of drug discovery, you do not need to disclose the AI method used to develop the drug/medicinal product for patentability at the EPO. But the product must be disclosed so it can be manufactured (i.e. enabled). Whilst in silico data may be useful before the EPO, wet lab data proving efficacy of the product will be necessary. Patent claims will need to be drafted as second medical use claims, and there must be plausibility for the claimed use (i.e. the compound linked to the second medical use actually being claimed).

AI and medical diagnosis and imaging

Another sector (or rather, several sectors!) seeing utilisation of AI is that of medical diagnosis and imaging. 

Similar considerations will apply as above, regarding whether the method in question would be best kept as a trade secret. Any method will need to be tethered to a technical effect – e.g. a diagnosis, production of images or some sort of process of data collection.

With patent applications in this area, there may be particular Sufficiency (Article 83 EPC) considerations. There is a need to ensure that the method can be carried out by a person of skill in the art reading the patent, so the method needs to be more than just method or code black boxes or an idea not backed up by a method that can be followed to achieve the claimed result. 

There may also be a further exemption to the patentability layer here, related to the exclusion of methods of diagnosis on the human/animal body. Methods need to use samples/data already collected from a human/animal body if diagnosis occurs as part of the method. Methods seeking patentability at the EPO cannot include steps of data collection on the human/animal body unless the method is non-invasive. Or, if the method includes collection from subject, the method cannot include the actual diagnosis, just the data collection. Consideration of these European requirements when drafting an application will lead to a smoother European prosecution.

Summary

Whilst there may be challenges to patenting these inventions in Europe, patents are regularly granted in these field. Patent applications can be drafted and prosecuted to account for the requirements.

As use of these tools becomes more commonplace, it will be particularly interesting to see how AI tools develop the consideration of “inventive step” before the EPO. Currently, just the use of AI instead of a human team is unlikely to grant an inventive step when something would be considered obvious. However, will it become harder to argue new products are inventive when powerful AI tools are perhaps capable of more than the human teams previously undertaking this? 

Please do get in touch with Liam Lawlor or any member of the Murgitroyd team to discuss any questions or inventions you might have in these fields.

For more in-depth insights and related information, be sure to check out our coverage on IPStars.