It was a full house for AI Patent Insights: securing future innovations at the AI Innovation Center this week. Bart Jan Niestadt from VO Patents & Trademarks and Twan Uijttewaal from the Netherlands Business Agency (RVO) came together for a knowledge session on intellectual property with a special focus on AI. Yesterday, we dove into Twan Uijttewaal’s speech. Today we focus on that of Bart Jan Niestadt.
Why is this relevant?
For start-ups, intellectual property rights are generally not a top priority. Yet it can be rewarding to think about it carefully, especially when operating in the field of artificial intelligence. Bart Jan Niestadt from VO Patents & Trademarks explains why.
When does an abstract concept become a technical advance? This question was at the heart of Bart Jan Niestadt’s exploration of artificial intelligence (AI) and its intersection with patent law. He and his audience delved into the complexities of patenting AI-related inventions. This was about differentiating what constitutes a patentable invention, particularly in the field of AI, which often straddles the fine line between technical progress and abstract concepts.
Niestadt begins by addressing the fundamental question of what defines an invention in the legal landscape, highlighting the approach of the European Patent Convention. The convention refrains from offering a definitive description of an invention, but instead provides guidelines emphasizing the need for technical novelty, inventiveness and industrial applicability. This framework paves the way for a nuanced discussion on the patentability of AI innovations, which often face the challenge of fitting into the technical category due to their inherent reliance on mathematical methods and computer programs . “It must therefore above all be new, inventive and industrially applicable,” explains Niestadt. “But they have several exceptions. The most important for us today are mathematical methods and computer programs, because this is also where artificial intelligence has its place.”
Crucial for AI
The speech highlights a key aspect of AI-related patents: the distinction between abstract concepts and their technical applications. Niestadt emphasizes a critical view: even if revolutionary discoveries in fields such as mathematics or natural phenomena are not patentable, their application to solve technical problems or create innovative tools can be. This principle is crucial for AI, where the application of machine learning models in technical fields turns abstract, unpatentable ideas into potentially patentable inventions.
For greater precision, Niestadt takes one of the exceptions, the discovery of natural phenomena. “If you are a brilliant researcher doing academic research on gravity and you make a revolutionary discovery, very important to the scientific world, you cannot patent it. Because what you have shown is a discovery that is already part of nature and our existence. This is not an invention. However, if you find a way to use this for technical purposes, for example if you create a new aircraft based on this principle, then suddenly you have an invention, because it is a technical application of your discovery. And you can find a patent application for it. This is how it works with each of these exceptions.
So the trick is to define this correctly and clearly to convince the patent office that you have a technical application for your invention. Niestadt is referring to the European Patent Office guidelines, which classify AI inventions as computer-implemented inventions. These guidelines emphasize that even though machine learning models are abstract in nature, their application in technical fields, such as image processing or signal processing, makes them patentable. This distinction is vital for innovators seeking to navigate the complex AI patent landscape.
Examples
Through concrete examples, Niestadt highlights the practical aspects of patenting AI technologies. He discusses patented innovations like an agricultural system using drones and AI for environmental monitoring, emphasizing the technical nature of these applications. Conversely, it presents the case of a rejected patent application for a method of determining cardiac output using AI, attributed to insufficient disclosure of the neural network training method, highlighting the importance of detailed technical descriptions in patent applications.
Niestadt’s exploration results in a crucial message for AI innovators: the path to patentability depends on demonstrating the technical application of AI models and solving technical problems. This approach aligns with legal patent requirements and ensures that AI innovations make a concrete contribution to technological progress. As AI continues to evolve and integrate into various industries, understanding and navigating the patent system becomes essential to protecting and fostering innovation in this dynamic field.
Other practices in the United States
Niestadt has another piece of advice for innovators who want to explore non-EU markets: “The rules may differ from country to country. You might consider filing in the US if you have something that isn’t patentable, and the US is often commercially relevant enough to get at least some protection if possible. In Europe, inventions that are clearly not considered technical may be valued differently in the United States. For example, the simple classification of a text based on its content is not patentable in Europe. If I read this, I automatically think of ChatGPT. We noticed that OpenAI does not in fact own any patents. But they have a handful of patents and applications in the United States. The patentability of this stuff there is just a little different than in Europe.
The first part of this short series was published yesterday. Read it here.
