AI and Patents: How to Avoid Inadvertently Giving Away Your Valuable AI-Generated Inventions

By Robert Plotkin and Cynthia Gilbert

This is a special series of articles about the impact of AI on patent protection from Blueshift IP: Software Patent Experts. This series is intended to be easy to understand and to focus on how to tailor AI patents to your business’ goals. This installment focuses on how artificial intelligence is making it more critical than ever before for innovative companies to choose between patenting their inventions and keeping them as trade secrets.

To explain this tradeoff, we’re going to use the example of Aesop’s fable, “The Goose That Laid Golden Eggs.” Do you remember hearing this story as a child? To refresh your memory, the story goes something like this. A farmer is checking his geese for eggs one morning when he finds, much to his surprise and delight, that one of the geese has laid a solid gold egg. He has no idea how or why this happened, but he confirms that, sure enough, the egg really is made out of solid gold and is worth many thousands of dollars. He sells the egg and is able to buy a new coop for his geese.

He wakes up the next morning and, once again, the same goose has laid another solid gold egg. This time the farmer buys a year’s worth of feed for all of his geese.

This pattern continues for a month, and each day the farmer sells the single egg that the goose has laid. He now has a comfortable savings and for the first time in his life is feeling some ease.

Yet at the same time, he is feeling some impatience. After all, the goose is laying only one egg a day. Although he knows that in a year this will yield 365 eggs, and even more after that, he thinks to himself, “Why wait a year for all those eggs? I have the goose right in front of me. If I cut the goose open I can have all of the eggs right now and not have to wait.”

The farmer, feeling proud of his bright idea, slaughters the goose and cuts it open, only to find that there are no eggs inside it. Now he has no goose and never sees another golden egg in his life.

If you’ve ever heard this story before, you’ve probably thought of it as a parable about the virtue of pride and the sin of greed. But have you ever thought of it as a story about the tradeoffs between keeping your inventions as trade secrets on one hand and patenting them on the other? We bet not!

But the fable of the goose that laid golden eggs illustrates perfectly the tension faced by today’s companies that are developing AI that can be used to generate new inventions. Maybe you are at one of those companies right now. If not, imagine that you are. Your company has developed and is constantly improving AI that can be used to generate designs for new airplane wings or drugs or even software. This AI has significant commercial value to your company. You might be inclined to patent your AI software in order to protect yourself against competitors copying that software or re-inventing it themselves and competing with you. That’s a very reasonable inclination to have and it may in fact be a wise business decision to patent your AI software for this reason.

But now think of your AI as a goose that lays golden eggs. The golden eggs in this analogy are all of the innovations that your AI can produce from now until many years into the future. Each of these eggs is an invention that may itself have significant commercial value. If you file a patent application for your AI, that patent application will be published, which will teach the world how your AI works. That may make it more difficult, or even impossible, to patent inventions that are created using your AI, depending on how much inventive effort is required to use your AI to create those inventions. So filing a patent application for your AI may result in you losing some or all of the value of the innovations that your AI can create. That is like killing the goose that laid the golden eggs.

And if you obtain a patent on your AI, eventually that patent will expire, and all of your competitors will be able to use your AI themselves to generate their own inventions. This is less like killing the goose that laid the golden eggs, and more like letting all of your neighbors clone that goose so that they can use those clones to lay their own golden eggs.

If these outcomes don’t sound good to you, what are your alternatives?  One of them is not to even attempt to patent your AI, and instead to keep it as a trade secret. That means not filing any patent applications for your AI, not publishing papers about it, and otherwise keeping it under lock and key and having everyone who has access to the AI sign nondisclosure agreements about it. The main benefit of this strategy is that you can potentially use your AI to churn out golden eggs forever in the form of new inventions, and you can then patent those inventions themselves individually as well. It’s like keeping your golden egg-laying goose alive and in secret to churn out golden eggs forever.

To play out this analogy even further, there are plenty of other potential business models. For example, imagine that you let a neighbor pay you to rent the golden egg-laying goose for a week, in exchange for which they get to keep the eggs that the goose lays during that week, and then they have to return the goose to you. This is analogous to granting another company a limited non-exclusive license to use your AI.

Our point here isn’t to argue that companies shouldn’t patent their AI technology. It’s to point out that reflexively patenting any new AI that you develop could result in you shooting yourself in the foot, and that instead it would be wise to carefully weigh the pros and cons of patenting your AI versus keeping your AI as a trade secret, and also carefully thinking through the business model that you will use to maximize the legal protection that you obtain for your AI and the profit that you generate from that AI in the long term.

In this sense, AI that can generate innovations is analogous to an employee who is an inventor within a company. Typically companies do not allow such employees to simultaneously work as inventors for competitors. Yet a company that publishes information about how its AI works, such as in the form of a published patent or patent application, is in some ways letting its AI work as an inventor for its competitors. This is something that all companies that develop AI today need to think carefully about as part of their IP strategy.

The ways in which AI is making inventors into super-inventors is just one of many impacts that AI is having on patent law and strategy. Thanks for reading this special four-part series on the impact of AI on patent law and strategy. If you are using AI to invent in your business and want to talk about obtaining strong, broad, and defensible patents for those inventions, please contact us directly at




Sign up to receive blog updates via email.

You Might Like