AI and Human Activity: Patently Indistinguishable

[1]

Overview

On September 5, 2025, U.S. Circuit Judge William Bryson handed down a decision on a patent infringement suit brought by ConGlobal Technologies against Roboflow Inc.[2] ConGlobal held three patents for systems where they utilize machine learning to manage the position of railcars on railways, U.S. Patent numbers 12,217,183 (‘183), 12,020,148 (‘148), and 12,254,439 (‘439), which they alleged were infringed upon by Roboflow.[3] Roboflow made a motion to dismiss the allegations, stating that the claims in the ‘183, ‘148, and ‘439 patents were unpatentable under 35 U.S.C § 101.[4] Judge Bryson granted the motion, holding the above patents as invalid due to the claims being directed to abstract ideas.[5]

Legal Background

Under 35 U.S.C. § 101, a patent must be a useful process, machine, manufacture, or composition of matter.[6] However, courts have historically excluded patents solely directed to laws of nature, natural phenomena, and abstract ideas (“judicial exceptions”).[7] The definition of an abstract idea has never been singularly articulated in any Circuit Court or the Supreme Court.[8] Instead, it is inferred from past case decisions and has enveloped ideas such as business practices and anything the court might consider to be a “building block of human ingenuity”.[9] Courts justify the exclusion of the judicial exceptions from patentability so that entire fields of human endeavor are not preempted by a single patent.[10] However, many innovations start as a judicial exception but constitute a marked improvement that might warrant protection.[11] The court in Alice laid out a two-part test to see if a claim directed to a judicial exception might still be patentable to protect such things.[12] Suppose a claim rests on the application of a judicial exception, but contains an inventive concept that allows it to amount to “significantly more” than the judicial exception itself. In that case, it may be eligible for patentability.[13] One of the underlying principles of the Alice test is to prevent people from getting patent protection on the aforementioned “building blocks” and their simple applications, which would hinder innovation.[14] However, material improvements upon these things are both encouraged and eligible for protection.[15]

Reasoning

In the case of ConGlobal’s patents, Judge Bryson found that they applied machine learning to a particular environment with no articulated inventive step added to it.[16] The Court reasoned that using machine learning to position trains based on their identity is analogous to using human observation to position objects, which is a fundamental human activity.[17] The lack of an inventive step fails the second part of the Alice test, making the claim ineligible for patent protection.[18] This decision follows a similar case decided earlier this year in the United States Court of Appeals for the Federal Circuit, where the claims were held ineligible when machine learning was applied to a data environment with no improvement to the model itself.[19]

The decision handed down by Judge Bryson is indicative of the current standard applied to machine learning and artificial intelligence-related patent applications and patents. Claims being directed to or involving machine learning or artificial intelligence are not inherently unpatentable.[20] They are, however, treated as an abstract idea and therefore are subject to the Alice test to determine their eligibility.[21] This is especially important in today’s evolving industries, where artificial intelligence and machine learning are being leveraged in new and beneficial ways.[22] Patent applicants operating in this growing area should heed this standard while drafting their claims or working on their inventions and be sure to be explicit in expressing how they’ve made improvements on the models they use in order to pass the Alice test.[23]

References

[1] Photo by Growtika, Stick and Ball Model Resembling a Brain, Unsplash (Jan 18, 2023), https://unsplash.com/photos/an-abstract-image-of-a-sphere-with-dots-and-lines-nGoCBxiaRO0

[2] See Theresa Schliep, Judge Axes Machine Learning Railyard Patents Under Alice, Law360 (Sep. 8, 2025), https://www.law360.com/ip/articles/2385439?nl_pk=8b61a0c2-107e-4802-af1c-fa34febfe7bf&utm_source=newsletter&utm_medium=email&utm_campaign=ip&utm_content=2025-09-09&read_main=1&nlsidx=0&nlaidx=10 (explaining the District of Delaware’s holding).

[3] See id.

[4] See All Terminal Servs., LLC v. Roboflow, Inc., Civil Action No. 25-476-WCB, 2025 LX 345262, at *1 (D. Del. Sep. 5, 2025) (describing patents in question and procedural history).

[5] See id.

[6] See 35 U.S.C. §101 (four categories of patentability).

[7] See E.g., Alice Corp. Pty. Ltd. v. CLS Bank Int'l, 573 U.S. 208, 227 (2014) (holding computer program that deals with financial transactions to be an abstract idea and thus unpatentable).

[8] See id.

[9] See id at 217.

[10] See Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66, 90 (2012) (holding method to be unpatentable since it did not add significantly more to law of nature).

[11] See id.

[12] See Alice Corp. Pty. Ltd., 573 U.S. at 217-18.

[13] See id. at 215.

[14] See id. at 208 (citing Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66, 89).

[15] See generally id.

[16] See All Terminal Servs., 2025 LX 345262, at *44.

[17] See Schliep, supra note 2.

[18] See All Terminal Servs., 2025 LX 345262, at *44; see also Alice Corp. Pty. Ltd., 573 U.S. at 217-18.

[19] See Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205, 1216 (Fed. Cir. 2025) (holding claims that apply established methods of machine learning to new data environment without any improvements to models themselves are ineligible for patent protection)

[20] See Akhtar & De Vellis, Patent Takeaways In Fed. Circ.’s 1st Machine Learning Rule, Law360 (May 13, 2025),  https://www.law360.com/articles/2339118?scroll=1&related=1 (explaining practical takeaways for practitioners).

[21] See id.

[22] See Anthony Cardillo, How Many Companies Use AI? (New 2025 Data), Exploding Topics (Nov. 3, 2025), https://explodingtopics.com/blog/companies-using-ai#activity-and-implementation-data (explaining statistics on ways companies are leveraging AI).

[23] See Akhtar & De Vellis, supra note 20.