<em>Recentive Analytics v. Fox Corp.</em>

Recentive Analytics Inc. v. Fox Corp., 124 F.4th 1205 (Fed. Circ. 2025)

In Recentive Analytics v. Fox Corp., the Federal Circuit held that using machine learning models in new data environments constitutes an abstract idea and lacks patent eligibility under 35 U.S.C. § 101, equating machine learning models to computers, or computer software, and signaling the need for advancement or innovation if novel applications of machine learning models are to be patentable.

On April 18, 2025, the United States Court of Appeals for the Federal Circuit handed down its decision in Recentive Analytics, Inc. v. Fox Corp, et al., affirming the lower court’s dismissal of Recentive’s patent infringement claim against Fox Corp. and its affiliates. See Recentive Analytics Inc. v. Fox Corp., 124 F.4th 1205 (Fed. Circ. 2025). This panel decision addressed a question of first impression and considered whether established machine learning methods could be patented when applied to a new field or environment.

Recentive Analytics Inc. is a Boston-based developer of machine learning platforms. Pitchbook, Recentive Analytics Overview https://shorturl.at/NXk5b (Last visited Mar. 8, 2026). Recentive specializes in predictive business optimization through big data analytics, notably for the National Football League (NFL). See id; see also Ken Bolson, How the N.F.L. Used Supercomputers to Set Its TV (and Streaming) Schedule, N.Y. Times (Sept. 7, 2023) https://shorturl.at/BUlcU. On the other hand, Fox Corporation and its affiliates are a national mass media company. Fox Corp., About Us, https://shorturl.at/ypNNS(Last visited Mar. 8, 2026). Part of Fox’s media network includes Fox Sports, which nationally broadcasts NFL football games. Seeid; see also Fox Sports, NFL, https://shorturl.at/bPS44 (Last visited Mar. 8, 2026).

Recentive claimed that Fox and its affiliates infringed on four of its patents. See Recentive,124 F.4th 1205.The first two patents consist of machine learning patents and applied artificial intelligence to optimize event schedules, while the second two patents are network map patents and applied artificial intelligence to optimize programming displayed in different geographic regions of the country. Id. at 1208–10. All four patents describe the same process of data collection, training/data analysis, and creation of an output. See id. at 1209–10. In addition, all four patents discuss how the machine learning techniques can be employed to any data sets, and how any generic computing equipment can be used to achieve similar results. Id.

At the district court, Fox moved to dismiss on the grounds that Recentive’s patents were ineligible and thus invalid. Id. at 1210. The district court granted Fox’s motion, holding that Recentive’s patents were ineligible under 35 U.S.C. § 101. Id. at 1211. Recentive appealed to the Federal Circuit and lost, with the Federal Court upholding the district court’s dismissal. Id. at 1215.

The Federal Circuit applied the two-step test adopted in Alice Corp. v. CLS Bank International and Mayo v. Prometheus to determine the validity of Recentive’s patents. Id. The first step of the Alice test is determining whether the patent contained an abstract idea. Alice found that using a computer program to manage an otherwise commonplace idea, a computerized trading platform that deals with financial transactions in which a third party settles obligations between two others so as to settlement eliminate risk, couldnot be patented. See Alice Corp. v. CLS Bank Intern., 573 U.S. 208 (2014). Therefore, under the Alice/Mayo test, algorithms and methods of computation could not be patented when adopted to a general principle. See Alice Corp., 573 U.S. 208 (2014); see alsoMayo Collaborative Services v. Prometheus Laboratories, Inc., 566 U.S. 66 (2012). If the patent is an abstract idea under the Alice/Mayo test, the patent must introduce “an element or combination of elements’ that transforms the claims into something ‘significantly more’ than a claim on the patent-ineligible concept itself.” SeeAlice Corp. 573 U.S. at 217–18.

In the first step, the Court found that all of Recentive’s four patents were abstract ideas. Recentive, 124 F.4th at 1212.The Court reviewed Recentive’s four patents and found that Recentive’s process of collecting data, learning or training with said data, and creating an output, are conventional techniques in data analysis. Id. The Court noted that Recentive itself acknowledged the patents cover machine learning applied to event management and geographic programming decisions, not machine learning itself. Id. at 1213. The Court concluded by reaffirming that “[the Court has] held the application of existing technology to a novel database does not create patent eligibility.” Id. at 1214.

The Court was not persuaded by Recentive’s argument that their implementation of machine learning tools was not generic or abstract because it “unearthed[ed] useful patterns that had previously been buried in the data, unrecognizable to humans.” Id. at 1213. The Court reiterated that in the context of computer-assisted tasks, computer-assisted methods that merely “speed up” human activity are ineligible under 35 U.S.C. § 101. Id. at 1214.

The Court found Recentive’s arguments for the Alice/Mayo Test’s second step to be wholly unpersuasive. The Court quotes Alice to say that to overcome the first step of the test, the patent must be more than stating the abstract idea, followed by “adding the words apply it.” Id. at 1215. The Court concluded that Recentive “plainly fail[ed] to identity anything that would transform the claimed idea into a patent-eligible application” (internal quotations omitted). Id.

Put simply, Recentive did not patent a new AI or LLM that can sort data and create an output. Instead, they patented the idea of taking pre-existing AIs or LLMs and applying them to sort certain kinds of data. This, according to the Federal Circuit, is an abstract idea ineligible under 35 U.S.C. § 101. See generally id.

‍ ‍Recentive suggests that AI, LLMs, and machine learning models are simply tools used to complete a task, and that their usage, for now, is equivalent to using any standard computer or computer software. This means that using an LLM to complete a human task, even if structured, does not satisfy 35 U.S.C. § 101. Recentive’s holding is not a novel reinterpretation of § 101 considering AI, it simply expands the Alice test to be inclusive of applications of AI. The Court also hints at the direction of AI-Patent litigation by focusing on the absence of a claimed technological improvement, as well as the failure of the claims at issue to “delineate steps through which the machine learning technology achieves an improvement.” Id. at 1213.

The ruling in Recentive does not, however, preclude machine learning models and other artificial intelligence from patent eligibility. A patent similar to Recentive’s for a MLM that requires specific hardware may have a better chance at eligibility. Similarly, if the hypothetical patent were to improve the usage of or with specific hardware, there may be a case for patent eligibility. With machine learning models becoming so pervasive in all aspects of society, it is only a matter of time before more patents are issued and challenged.