29 February 2024

Common life science patentability issues working their way into AI applications

Artificial Intelligence (AI) is a rapidly growing field that has been making significant contributions to various industries, including the pharmaceutical industry. In this article, we will discuss the use of AI in drug discovery and the patentability of AI-based inventions, as well as introduce the concepts of plausibility and sufficiency and how they relate to such inventions.

AI is a broad term that refers to the ability of machines to perform tasks that would typically require human intelligence. AI is used in various applications, including natural language processing, image recognition, and drug discovery. In the pharmaceutical industry, AI is used both to discover new drugs and new uses for existing drugs, for example. This takes advantage of the ability of AI to analyse large amounts of data and identify patterns that humans may not be able to detect.

The patentability of AI-based inventions is a complex issue. For example, the patentability of a new drug discovered using AI is distinct from the patentability of the AI method itself. The drug may be patentable if it meets the requirements of novelty, non-obviousness, and industrial applicability. However, the patentability of the AI method itself may be more complicated due to the nuances of patent eligibility issues around both mathematical methods and software.

Plausibility has previously been relevant mainly to applications in the chemical and life sciences technology area, however, recent case law has highlighted its importance in AI-based inventions. Plausibility refers to the requirement that a patent application must provide a plausible technical solution to a technical problem. Likewise, sufficiency objections can be raised for all types of patent applications at the European Patent Office (EPO). Sufficiency refers to the requirement that a patent application must provide enough information to enable a person skilled in the art to carry out the invention. Such objections are common in applications for pharmaceutical or biotechnology applications, and experimental data is used to address them (see more here).

Similar issues can sometimes also arise for AI inventions, especially for AI cases in the life sciences field. This may, for example, arise due to a lack of sufficient detail in the description of the model, or the training process or the input data used to train the model (e.g. as in T0161/18 and T2803/18). It is therefore important to include as much detail of the model, the training process, training data, and the output information as is practical. This varies on a case-by-case basis; some details may be irrelevant for some models whereas they are essential to others. Lack of sufficient disclosure is a very difficult objection to address at the EPO once it has been raised, so it is imperative that a sufficient level of detail is included in patent applications from filing.

Decisions such as G2/21 and those highlighted above have led the EPO to update their guidelines (e.g., G-II 3.1.1) to state that “the technical effect is dependent on particular characteristics of the training dataset used, those characteristics that are required to reproduce the technical effect must be disclosed unless the skilled person can determine them without undue burden using common general knowledge. However, in general, there is no need to disclose the specific training dataset itself”. In light of this update to the guidelines and the associated recent case law, it is important that care is taken when drafting an application to ensure that enough detail regarding the training data is included to allow the skilled person to reproduce the technical effect across the whole scope of the claimed invention.

In addition to protecting a method, it is often equally as important to protect the method outputs. An example of this is if an AI model is used to find biomarkers, we would advise considering a claim to the biomarkers themselves. Such product type claims are often easier to defend, define, and detect infringement. While protecting both the model and the outputs is often the most desirable outcome, under certain circumstances, such an approach may also lend itself to the possibility of keeping the AI method as a trade secret. However, this approach needs to be carefully considered, and we advise consulting with one of our patent attorneys before deciding on your strategy.

Finally, we recommend speaking to our team of cross-disciplinary experts to discuss your commercially-focused IP strategy. Our team can help you navigate the complex issues surrounding AI-based inventions, in particular when applied to the life sciences technology area, and help you develop a patent and IP strategy that meets your needs.