Hottest Patent Term Of 2026? SMED.

Every year has its "it" term.
In 2025, the crown belonged to AI, and rightfully so. AI dominated the headlines, flooded the USPTO's dockets, and triggered more §101 rejections than any examiner would care to admit. If you worked in tech, you had to talk about AI. If you worked in patent law, you likely had to explain why simply applying AI to an existing technology is not patentable.

As a primer, a §101 rejection is basically the USPTO telling you, "Interesting idea, but the law says you can't patent that," usually because the claimed invention is an abstract idea. For AI inventions, this often means the USPTO argues that the model is simply performing a function that could be carried out by the human mind, i.e., "a mental process."

2026 is shaping up differently. This year's breakout term will not come from Silicon Valley, and it will not be something your non-legal friends casually drop into conversation, unless you are hanging out at the bar after an IP conference (and let me tell you, patent attorneys know how to party). Instead, the hottest term of 2026 is about to come from deep within the halls of the USPTO, courtesy of a December 4, 2025, memo with the energy of "I can't believe no one remembered this existed."

Ladies and gentlemen...2026 belongs to SMED. Yes, Subject Matter Eligibility Declarations.

A tool so underused and under-discussed that half the practitioners who read the memo had the same reaction: "Wait... we can file evidence for §101? We've been fighting this with words alone?"

Like the Brussels sprouts resurgence of the early 2000s, suddenly SMED is having its main-character moment.

So What Is a SMED, Exactly?

An SMED is a Rule 132 declaration that allows applicants to submit evidence, rather than just an attorney argument, in support of patent eligibility.

That evidence can include experimental results, performance data, architectural explanations, expert testimony, state-of-the-art background, and inventor declarations explaining why the claimed innovation was not routine or conventional.

In other words, an SMED gives applicants a way to show, with facts, why their invention improves technology.

A simple example

Suppose your claims involve a neural network with distributed layers performing gradient-adaptive routing and multi-modal fusion (meaning the system breaks complex data into parts, routes each part to the best tool for the job, and then blends everything back together, essentially processing thousands of conversations, images, and signals at once). An examiner may assert that the steps "can be performed mentally," such that the claims are rejected for being directed to an abstract idea. Under an SMED approach, your inventor can submit a declaration explaining why a human mind cannot perform the network's operations: humans cannot simultaneously compute multi-million-parameter tensors, apply real-time backpropagation, perform vectorized attention weighting, or route inputs dynamically across expert layers.

The shift began with the Desjardins decision, now precedential, in which the USPTO formally recognized improvements to machine-learning architectures as technological advancements, not abstract ideas. Reduced storage needs, lower system complexity, and the avoidance of catastrophic forgetting are technological improvements. They are eligible. They are patentable.

The memorandum from Director John Squires reminded the examining corps, and indirectly practitioners, that applicants do not need to rely solely on attorney argument to overcome §101 rejections. They can submit declarations. They can bring in experts. They can introduce objective data, performance comparisons, architecture breakdowns, or state-of-the-art testimony.

In other words, you can actually prove your invention is eligible. An SMED may be used to prove that the invention is not directed to a mental process or that it improves a neural network.

Shocking, I know.

SMED transforms the eligibility conversation from metaphysics ("But is this really an abstract idea?") into something grounded, factual, and technical. It gives applicants a way to connect the dots between what an invention actually does and what an examiner may be glossing over. For AI, robotics, animation systems, signal processing, fintech, and any architecture-driven technology, SMED may be the star ingredient.

The truth is, SMED has been hiding in plain sight for decades—a quiet Rule 132 declaration. An optional filing that most practitioners never consider. But like Brendan Fraser's resurgence, all it needed was the right moment, and now that moment is here.

So here we are.
2025 was the year AI took over the world.
2026 will be the year patent lawyers start asking, with increasing frequency and only a hint of dramatic flair:

"Should we just SMED this?"

More often than not, yes.

Welcome to the Year of SMED.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.