Artificial intelligence (AI) is reshaping numerous technological fields, from medicine to manufacturing to finance. This technological revolution raises intricate questions regarding intellectual property, notably concerning the patentability of AI.
Patents have traditionally been granted for human-made inventions, but what about inventions generated by machine learning algorithms? Nothing new according to the European Patent Office: a computer-implemented invention can be patented if it meets the traditional patentability criteria: novelty, inventive step, and industrial applicability.
For an AI-related invention to be considered novel, it must not be anticipated. If an AI technique has already been described in a prior document, the invention cannot be considered novel. As AI is evolving rapidly, our experts will conduct a thorough examination of the prior art.
Inventive step, i.e. a non-obvious advancement over the existing state of the art, is another crucial criterion. In other words, our experts need to ascertain whether AI has genuinely brought significant innovation.
Industrial applicability means that the invention cannot be purely theoretical but must have a concrete application. Many AI applications, such as image recognition, data analysis, and robotics, fulfill this criterion, making patentability of AI plausible in many cases.
Our experts will need to determine the technical solution to a technical problem brought by the AI-related invention. Technical features may include technical data that can be measured used as input in an AI model (such as temperature, image as light, blood pressure, etc), technical data processed into technical data as input in a AI model (light as pixel information), or non-technical data transformed into technical data (transcription of text related to technical data such as clinical notes).
AI systems for example can be trained to analyze medical images like scans or X-rays to detect anomalies or diseases. In this case, AI provides a technical solution to a technical problem by enhancing the accuracy and speed of medical diagnosis. Such an application of AI would be considered patentable as it pertains to a technical goal.
However, not all uses of AI necessarily fulfill the requirement of a technical solution to a technical problem. For instance, a machine learning algorithm developed to recommend products to users based on their purchase history does not necessarily solve a specific technical problem but rather a commercial or marketing one.
In a nutshell: even if abstract algorithms are generally not patentable, and thanks to a close collaboration with our patent experts, to ensure that the patent claims are carefully drafted, to be sufficiently specific for patentability while avoiding excessive restriction, patenting your AI software is possible!
With AI continuing to push the boundaries of technological innovation, patentability of AI is likely to remain an important and fascinating topic for years to come.
Authors: Olivier Nicolle, European Patent Attorney and Olivier Laidebeur, Patent Attorney
If you have any questions about your intellectual property, your brand or your IP assets in general, don’t hesitate to contact the experts at Laidebeur & Partners.
