About the inventor, the prompt and the patent.
Now that artificial intelligence has become an indispensable assistant in just about all sectors, we at Oryon are increasingly getting AI-related questions. The most frequently asked question is: can I get patent protection for the tool that I have (partly) created with prompts?
We consulted our colleague Michaël Beck, European patent attorney and co-author of the book ‘Patents in Belgium – a practical guide’.
Let’s start with another frequently asked question: can an AI agent be an inventor under patent law?
Case law is very categorical in this: no, that is not possible. Someone put an invention devised by an AI agent into a patent application, with the agent’s name as the inventor on it. That person did so with the aim of forcing the European and American Patent Offices to take a position on this. He therefore clearly indicated that it was an artificially developed invention. That application was rejected on the ground that the inventor is deemed to be a natural person under patent law.
“According to patent law, an inventor must be a natural person.”
Can a natural person who makes an invention with the help of AI get patent protection for it?
To be honest, legal science today is not yet completely sure how to deal with cases developed by AI. It is not easy to determine when there is sufficient human input to say that that person actually has the rights to that invention.
You have to know that the entire patent system, as we have known it for more than two hundred years in continental Europe, is intended to encourage innovation. In return for this effort, the inventor is given a temporary monopoly. That logic does not work so well if the entire intellectual process is replaced by the push of a button. In practice, of course, it’s not just one push of a button, but an interactive, iterative process where you start with a prompt, get suggestions from AI, and build on that. That is not quite the intellectual process we envisioned when patent law was drafted, but it is also not the other extreme of one push of a button.
Where exactly you should draw the line between ‘mainly a human activity’ or ‘mainly a machine activity’ is not clear. There is currently no legislation that draws this line. So for the time being, we have to look at it on a case-by-case basis and follow up on which cases are disputed and what is pronounced by judges.
If someone comes to Oryon for the patent application of an invention that was (partially) written with the help of AI, can you help them?
As with other inventions, we must first find out whether the invention meets the conditions for patenting. To begin with, it must be a technical solution to a technical problem. As soon as a computer or an AI agent appears in the patent claim, the claimed matter is technical. In addition, the European Patent Office examines whether the technical invention is also new, inventive and industrially applicable .
The patent granting authority will only accept that it is an inventive step if a step has been taken that is not obvious to a skilled person, given what is already known and the problem that had to be solved. You could argue that the application of AI has become a matter of course for the professional today. It is evident in many sectors to apply well-known AI models, whose capabilities are well known, to problems to which AI had not been applied before. So if the creation of the inventor only consists of solving the problem that we have always solved with pen and paper or with research in a database, now with AI, then in most cases the threshold of inventive step will not be reached.
“What counts for patent law is: does it have that inventive step ? It certainly happens that things that are innovative or even groundbreaking in the eyes of the inventor do not reach that threshold of inventive step in patent law.”
OK, but suppose it is indeed an inventive step, made with the help of AI, is that invention patentable?
In that case, it does not matter whether you have programmed the invention completely yourself, or whether you have used a tool to help with programming. We are not discussing the line between coming up with your own ideas and letting them think about it by AI, because it is purely about implementation. If you have already taken the step of coming up yourself, you can use all the tools you want to get from your concept to a working product.
“You can use all the tools you want to turn your self-invented concept into a working product, including artificial intelligence.”
But without a patent, anyone can use the invention?
There is also such a thing as trade secrets. Know-how has a certain degree of protection in law. This is about information that you have as a company, that you have made an effort to gather, and that has an economic value. For example: the way in which a certain AI system is trained and which is not visible to the end user is a confidential trade secret of the provider of that system. If we use a commercial chatbot today, think of ChatGPT or something, we only see the user interface, we can ask ChatGPT questions and get answers. But we don’t see what has preceded in terms of training and configuration. This part can be regarded as a trade secret that is protected as such against copying, albeit only under specific conditions. After all, you are only protected against abuse by parties who have unlawfully gained access to that secret information. So someone who can reproduce the exact same configuration with their own efforts and create a system that gives the same answers as ChatGPT is not doing anything wrong, as long as that person has not used the secret information for this.
“If you want to be sure that no one copies your invention, a patent offers more protection than just the trade secret.”
Do you have to substantiate in the patent application how you used AI to arrive at your invention?
If you want to be able to demonstrate that the use of AI makes things inventive, then you will indeed have to describe in the patent application how you used that AI. A skilled person who reads that patent must also be able to reproduce that invention. One of the concepts of patent law is that when you make an invention public, you get a temporary monopoly for it, but at the end of that monopoly the rest of the community must be able to enjoy that invention. And that is only possible if you describe in your patent application in sufficient detail how that invention works. You can’t say: ‘I want the monopoly, but I’ll keep the secret sauce to myself’.
Sounds logical. One last question: does the use of AI not pose dangers in terms of confidentiality and novelty?
That’s definitely something to be careful about. An invention is only new if it has not been publicly announced. So if you use an AI tool that offers no guarantees in terms of confidentiality, then you are already making your invention known and therefore jeopardizing your chances of patent protection. This is on the assumption that a third party could prove afterwards that you made that announcement. For example, we see more and more people who have already prepared their patent application with the help of AI. I would rather advise against that. Not only because we apply our own, in-depth procedure for drafting and filing patents, but also because of confidentiality.
Interesting information all. Thank you for this clear explanation Michael!
Do you still have questions about the patentability of your idea? Don’t hesitate to contact Oryon’s intellectual property experts.
