WHY I CAN’T FILE YOUR AI-PRODUCED PATENT APPLICATION

You have an idea. You work with your AI to develop the idea until it looks great to you. The AI then asks if you want it to write a patent application. “Wonderful”, you think, “If the AI writes the patent application, I won’t have to pay a lot of money to my patent attorney for drafting an application!”

When you send the AI’s document to your patent attorneys, they still want to charge you for their review and/or their full drafting. WHY?

Well, it could be grouchiness or selfishness – the patent attorneys are not happy that clients want AI to replace them.

But based on the AI-generated documents we’ve been receiving, it’s more than that. The AI-generated documents may be styled like patent applications, but the kind of information being disclosed is very similar to what we’ve been receiving from clients, even long-term clients, for years. And we’ve had to rewrite those disclosures to “convert” them into patent applications.

Conversion is the wrong term. Client documents generally discuss the user-facing aspects of their products and rarely focus on the inventive elements that are often deep within the product. Most products have an inventive core wrapped within standard product elements (like a housing or, for software, lots of standard elements like login, security, interface, etc.). A patent application concentrates on the inventive core and its implementation. Thus, our first “conversion” job is to define the “inventive step” within the product and then structure the application to bring it out.

However, since everyone knows you want broad claims, they assume their description should be broad as well. While this is true for part of the description, it is not true for the whole. In patent law, there is an “enablement” requirement – you have to provide a description of how to implement the invention that you’ve described broadly (i.e., not how to implement the product but how to implement the inventive step aspect we have defined). Without that enablement, your application lacks a proper basis for the claim you are making.

As a result, the documents we receive, AI-generated or not, are only appropriate as invention disclosures. They are missing too many important elements to be filed as full patent applications.

Why do AI-generated patent applications seem proper? Since they were trained on a vast corpus of prior patents, AIs know the structure of published patent applications, so the documents they produce look like patent applications, but their content is misdirected and usually incomplete.

There’s another surprising problem. The AI-generated disclosure documents require just as much correction as the documents we’ve received in the past! They are better in some ways, but we still need to ensure that concepts discussed in the text are shown in the drawings, and vice versa, that elements or steps in the drawings are discussed in the text, etc. A new problem specific to AI-generated documents is that we need to check for hallucinations in the list of cited articles or patents!

This matters because you are not allowed to add “new matter” to your spec. Ever! You are not allowed to use new terms or bring in new facts that weren’t already in or implied in the spec. Some patent offices, like the European Patent Office, are extremely strict about this. Other countries are more lenient about what might be implied in the spec, but none of them allow new facts.

So, if you didn’t put a fact into the spec to begin with, you can’t add it later, when you might need it to overcome a piece of prior art. And, even if you could add a fact, you usually can’t for a different reason –you have already made your wonderful product publicly available in some way, so the fact is no longer ‘new’ to the world.

This means that your spec has to describe your invention at many levels – from the broadest down to the implementation details. Any one of those details may be needed during examination to overcome rejections.

Therefore, we patent attorneys work hard to make sure that the spec is consistent and accurate and that it contains all the information in it that you might need at some point during the next 20 years of the patent application’s life. We need to consider how we present the invention, which details we need, and how your invention may change over time.

This is the talent a patent attorney brings to your application: elucidating what is inventive in what you did and presenting it via the drawings and the text that discuss them. If you will, we patent attorneys are skilled at “disclosure assistance”, which involves designing the presentation of the invention and its implementation in text and clear drawings.

Your material, AI-generated or otherwise, is still very important, as it is usually the first technical writeup of what you or your team has envisioned. The process of writing it often helps you work out some of the kinks in your product. Importantly, it provides us with a wealth of information about your product, including context for the product or invention, your terms, marketing or otherwise, and details of what you consider important. Much of this we will use in the patent application. Having this document makes it much easier for us to write the patent application. We just want it in your language – we don’t need it to look like a patent application.