Recentive Analytics, Inc. v. Fox Corp. A Case of First Impression on Machine Learning and § 101: An AI Preview?

Addressing a question of first impression, the Federal Circuit recently clarified limits on patent eligibility for inventions related to machine learning.  

The panel held that patents directed to using generic machine learning techniques in a particular environment, without claiming a specific method to improve the machine learning techniques, fail under the two-step Alice test.

Background.  Under the Supreme Court’s precedent in Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208 (2014), courts undertake a two-step test to determine patent eligibility.  In step one, the question is whether the patent claims are directed to a patent-ineligible concept, such as an abstract idea.  Id. at 217-218.  If so, in step two, the question is whether the claim elements, together or individually, contain an “inventive concept” sufficient to “transform the nature of the claim into a patent-eligible application.”  Id. at 217.  Recentive is the first time that the Federal Circuit has applied the two-step Alice framework to machine learning patents.

The Recentive decision has implications for both patent owners and accused infringers.  The decision highlights the importance for companies to evaluate the risks that potential ineligibility challenges pose to their patent portfolios, especially for patents relating to or using machine learning techniques.  Companies should implement forward-thinking strategies to navigate such challenges.  The decision also provides insights for parties litigating 35 U.S.C. § 101 in the machine learning, artificial intelligence, and related spaces.

I. Recentive Analytics, Inc. v. Fox Corp.  In its April 18, 2025 decision, the panel seized on the opportunity to address a question of first impression: “whether claims that do no more than apply established methods of machine learning to a new data environment are patent eligible.”[1]  The decision clarifies that such claims are ineligible under § 101.  The panel affirmed a Delaware district court’s dismissal of Recentive’s infringement claims asserting patents using machine learning to generate network maps and schedules for television broadcasts and live events.

At issue were two sets of patents that used machine learning.[2]  The “Machine Learning Training” patents were directed to collecting event parameter and target feature data, iteratively training a machine learning model based on relationships within the data, outputting an optimized schedule, and generating new schedules based on changes to the data inputs.[3]  The “Network Map” patents similarly were directed to collecting broadcast schedule data, analyzing that data to create a network map, updating the network map based on changes to the data inputs, and using the network map to optimize live events.[4] 

These patents applied machine learning techniques to specific contexts:  event scheduling and network map creation.  But the patents did “no more than apply established methods of machine learning to a new data environment,” without improving those techniques.[5]

• At step one of the Alice inquiry, the panel found that Recentive’s patents were “directed to ineligible, abstract subject matter,” noting that the patents do not “claim[] machine learning itself” and “rely on the use of generic machine learning technology in carrying out the claimed methods for generating event schedules and network maps.”[6]  Critically, the claims “d[id] not delineate steps through which the machine learning technology achieves an improvement.”[7]  By contrast, the panel suggested that claims that recite specific “improvements to the machine learning models to be applied” may be patent eligible.[8]
• At step two of the Alice inquiry, the panel determined that the claims failed to “transform the claimed abstract idea into a patent-eligible application” and instead “claim[ed] the abstract idea itself.” [9]  Specifically, the panel explained that “[t]ransforming the nature of a claim into a patent-eligible application requires more than simply stating the abstract idea while adding the words ‘apply it,’” and rejected Recentive’s argument that “‘using machine learning to dynamically generate optimized maps and schedules based on real-time data and update them based on changing conditions’” was inventive.[10]

II. Key Implications.  This decision has significant implications for companies seeking to protect their machine learning inventions.  Such inventions may be patent eligible only if the patentee is willing to disclose the specific algorithm used to improve machine learning in the patent, itself.  Companies should consider whether to do so or instead to rely on trade secret protection for generic use of machine learning in new environments.  

The decision also has implications for patent owners and accused infringers litigating § 101 challenges.  For instance, the order focused 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.”[11]  That reasoning provides guideposts on how such disputes might be framed. 

Notably, the panel recognized that “[m]achine learning is a burgeoning and increasingly important field and may lead to patent-eligible improvements in technology” and held “only that patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101.”[12]  Future cases will shape the contours of patent eligibility in the machine learning and potentially artificial intelligence contexts.

[1] Recentive Analytics, Inc. v. Fox Corp. et al., No. 2023-2437, slip op. at 10 (Fed. Cir. Apr. 18, 2025).

[2] Id. at 2-3.

[3] Id. at 3-4.

[4] Id. at 5-6.

[5] Id. at 10.

[6] Id. at 11.

[7] Id. at 13.

[8] Id. at 18.

[9] Id. at 16-17 (internal quotations omitted).

[10] Id.

[11] Id. at 13.

[12] Id. at 18.