Patenting Machine-Learning Inventions After the Latest Section 101 Guidance

Home / Insights / Intellectual Property

Intellectual PropertyPatentArticle

Published: April 29, 2026
Author: Casey Scott McKay
Reading time: 2 min

Subject-matter eligibility remains the central battleground for AI inventions. This article explains how we draft machine-learning claims to survive both examination and litigation under the latest Section 101 guidance. The drafting lesson is consistent: tie the claimed method to a specific technical problem and a measurable technical effect. We also explain why a deep specification is essential to a flexible continuation strategy. The difference between a dead end and a durable portfolio is usually made at the drafting stage.

Eligibility Is Still the Make-or-Break Issue

Subject-matter eligibility continues to be the make-or-break issue for machine-learning patents. The updated guidance gives examiners more structured criteria, but the burden is still on applicants to show a concrete technical improvement.

The drafting lesson is consistent: tie the claimed method to a specific technical problem and a measurable technical effect. Abstract descriptions of 'training a model' invite rejection; claims grounded in architecture, data pipelines, or hardware utilization fare far better.

Draft for the Portfolio, Not Just the First Claim

We also counsel clients to build a specification deep enough to support continuation strategy. The first claim set is rarely the last, and a rich disclosure is the difference between a flexible portfolio and a dead end.

Section 101 Subject Matter Eligibility Machine Learning Artificial Intelligence Patent Prosecution Software Patents Technical Improvement Claim Drafting USPTO Guidance Continuation Strategy Specification Alice