Is it possible to patent Artificial Intelligence?

Artificial intelligence and machine learning are applied across numerous technological fields. However, the protection of inventions in the IT sector is often uncertain, raising doubts among inventors.

How does the patentability test for Artificial Intelligence (AI) work according to the European Patent Office (EPO) guidelines? The EPO applies the same framework to AI as it does to all computer-implemented inventions, known as the "Two-Hurdle" approach.

The First Hurdle concerns technical character (Art. 52 EPC).

The patent claim—the part that defines the subject matter of the invention and the scope of protection—must have a technical character. In practice, it is generally sufficient for the claim to refer to a technical means, such as a computer, a neural network, or a processor. For example, the use of a neural network in a heart-monitoring device to detect arrhythmias constitutes a technical contribution.

The Second Hurdle concerns inventive step (The "Comvik" Approach).

According to this approach, only the distinguishing features that contribute to the technical character of the invention—namely, those that provide a contribution in a technical field and solve a technical problem—can be taken into consideration.
The EPO examines the mathematical steps of the AI algorithm and asks: do they contribute to a technical solution to a technical problem?

• If the answer is YES: the technical contribution is recognised—for example, when AI processes medical images to identify tumours or an algorithm controls the movement of a robotic arm. In these cases, the mathematical steps are taken into account for the purposes of the inventive step.
• If the answer is NO: the technical contribution is not recognised—for example, when AI predicts financial market trends (a business method) or an algorithm classifies text for linguistic purposes. In these cases, the mathematical steps are simply ignored.

In summary, the EPO treats AI as a mathematical method, patentable only when applied to a specific technical field. For researchers and university spin-offs, the key lies not in the algorithm itself, but in the ability to demonstrate how it solves a concrete technical problem.

USI Transfer, the University's Technology Transfer Office, is available for support or clarification. Please write to [email protected]

Further reading
Specific section (Part G-II, 3.3.1) of the EPO Guidelines.