AI Patents After Desjardins: Keeping American Leading in Tech?

‍There has been what might be an important change in the patent landscape regarding inventions based on Artificial Intelligence and Machine Learning. Let’s take a look. This could be big. Or not. It depends.

Some Background

Let’s start with the basics.

A patent begins with an application filed at the Patent Office. The application has two core parts. The specification lays out the invention, that is, what it is, how it works, and what problem it solves. The claims define the boundaries of the invention. Think of the claims like a real estate description. They mark out what the inventor does and does not own. The chains define whether another device, process, or system is “trespassing” on — “infringing” in patent parlance — the patent. If a patent is being infringed, the owner can apply to a court to seek damages and stop the infringer from making, using, or selling the infringing device, process, or system. Millions of dollars can be involved.

And companies are usually more likely to invest in research and development of technology they can patent, giving them exclusive rights. If a competitor can simply copy the invention and sell it, investment in the technology may not be attractive.

Once filed, the application is examined. The United States Patent Office employs patent examiners to review whether the invention claimed in a patent application is new (novel), whether it would have been obvious (just a minor, predictable change to a known invention), and whether the claims are supported by the specification.

The Alice/Section 101 Test

And, most important to today’s discussion, is the question that has caused plenty of trouble in recent years: Is this patentable subject matter?

That question comes from the Patent Act section 101 and how it was interpreted in the Supreme Court’s decision in Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (201). The rule applied there was that you can’t patent abstract ideas. The Court set up a two-step test. First, is the claim directed to an abstract idea? Second, if it is, does it include something more—an ‘inventive concept’ that turns it into a real application?

That test has been hard to apply. The first step seems more philosophical than technical in many cases. In applying the test, courts and patent examiners tend to sort claims into categories—mathematical concepts, mental processes, or methods of organizing human activity. Once a claim lands in one of those categories, it is often doomed.

AI inventions have had a particularly tough time. Many AI systems can be described as math. Others look like automated decision-making—and that looks like a mental process. That’s the point of AI, right? Still others resemble ways of organizing information or activity. The result has been examiners rejecting many AI-related patent applications. To survive, applicants have had to show the ‘something more’ from Alice—a concrete, non-routine technological element. That has not always been easy to show.

Enter Ex parte Dejardins

Now let’s takes a look at Ex parte Dejardins, Appeal No. 2024-000567 (P.T.A.B. Sept. 26, 2025) (precedential).

This was not a court decision. It came from the Patent Trial and Appeal Board. Think of it as the Patent Office’s own internal administrative court. PTAB decisions do not bind courts. But when the Patent Office designates a decision as precedential — as it did for Dejardins — it becomes highly persuasive inside the agency. Examiners follow it, and PTAB panels follow it. And, since courts consider the Patent Office to be the expert on patent law, PTAB decisions are often persuasive. Not as persuasive as Federal Circuit decisions mind you, but still persuasive.

Desjardins shifts the focus for AI-related inventions. Instead of focusing primarily on whether an AI invention looks abstract in the first place, it focused more carefully on whether the invention reflects a technological improvement. Does it make the computer work better, faster, or more efficiently? Does it solve a real technical problem in computing or machine learning?

Approaching the issue this way is based on a broader point made in Dejardins. Here is what the PTAB said:

“Under a charitable view, the over-broad reasoning of the original panel below is perhaps understandable given the confusing nature of existing § 101 jurisprudence, but troubling, because this case highlights what is atstake. Categorically excluding AI innovations from patent protection in the United States jeopardizes America's leadership in this critical emerging technology. Yet, under the panel's reasoning, many AI innovations are potentially unpatentable-even if they are adequately described and nonobvious-because the panel essentially equated any machine learning with an unpatentable "algorithm" and the remaining additional elements as "generic computer components," without adequate explanation. Dec. 24. Examiners and panels should not evaluate claims at such a high level of generality. “(Emphasis added.)

The concern about America’s role in AI leadership apparently caused the Patent Office to move fast. Shortly after Desjardins, the Patent Office announced it was updating the Manual of Patent Examining Procedure—the MPEP—to incorporate the focus-on-the-technological- improvement approach. The MPEP is not law. But in practice, it is the blueprint for patent examination. Examiners use it to decide patentability, and practitioners draft to it.

At about the same time the office focused the patentability question on technology, it also issued guidance on Subject Matter Eligibility Declarations—SMEDs. The change allowed applicants to submit evidence showing that their invention improves technology in a meaningful way.

AI inventions

This all bodes well for AI inventions. The decision was written by the Director of the Patent Office himself, signaling what one might call a change in attitude to AI patents. While still applying Alice, the Director thought too much emphasis had been given to the first step, with the second step getting short shrift. No more will that be so. And to emphasize the point, the “patent prosecution bible” – the MPEP – will be revised consistent with that approach along with treating technologic proof as evidence, not just argument.

What about the courts?

AI patents will still have to survive court review under the Alice framework if challenged. Still, Patent Office guidance will shape how applications are examined and how records are made.

And so . . .

This change is likely to result in more AI-related patents being issued. But will it save America’s leadership in AI technology? It depends.

This change is bound to result in more AI patents. But it’s up to the courts whether they will be valid. As noted, the courts will apply Alice and may or may not follow the Desjardins approach to the Alice test.

Meanwhile, there are ongoing attempts to change criteria for patentability legislatively and to get the Supreme Court to take another look at the current interpretations of Alice. Those efforts may moot Dejardins’ effect.

And patent office directors can change. A new director may take a different approach.

So, we won’t know how this will all end up for years. But we do know that, for now, the Patent Office has changed its approach to AI-related patents in a way that is favorable to their issuance. And applicants have a new tool for proving eligibility.