On April 18, 2025, the United States Court of Appeals for the Federal Circuit ("Federal Circuit") issued a significant decision in Recentive Analytics, Inc. v. Fox Corp., affirming dismissal, by the District Court of Delaware, of patent infringement claims brought by Recentive Analytics, Inc. (“Recentive”) against Fox Corp. and affiliates (collectively, "Fox").  The case centered on four patents owned by Recentive—U.S. Patent Nos. 10,911,811, 10,958,957, 11,386,367, and 11,537,960—which purported to cover the use of machine learning for generating event schedules and network maps, particularly in the context of television broadcasts and live events. 

The patents generally fell into two categories:

• Machine Learning Training Patents (’367 and ’960): These patents described methods for dynamically generating event schedules by training machine learning models on historical data and updating schedules in real time based on changing parameters. 
• Network Map Patents (’811 and ’957): These patents described methods for generating and updating network maps for broadcasters, using machine learning to optimize outcomes such as television ratings. 

According to the Court, each of the patents ultimately relied on “generic” machine learning techniques (e.g., neural networks, support vector machines, decision trees) implemented on standard computing hardware.  

Fox moved to dismiss the complaint in the District of Delaware, arguing that the patents were ineligible under 35 U.S.C. § 101 because they were directed to abstract ideas and lacked an inventive concept.  The District Court agreed, dismissed the case, and denied Recentive leave to amend its complaint.  Recentive appealed to the Federal Circuit.  

Federal Circuit Holding

The Federal Circuit affirmed the district court’s dismissal, applying the two-step framework for patent eligibility established in Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014), as set forth below. In particular, the Federal Circuit noted that this case presented a question of first impression: “whether claims that do no more than apply established methods of machine learning to a new daata environment are patent eligible.” 

Step One: Directed to an Abstract Idea?

The Federal Circuit agreed with the District Court that the claims were directed to the abstract ideas of producing event schedules and network maps using known mathematical and computational techniques. The patents did not claim any improvement to machine learning technology itself (or to the underlying computer technologies being used, but rather the application of generic machine learning methods to the fields of event scheduling and network mapping. 

The court emphasized that the patents did not disclose any specific implementation or technological improvement in machine learning.  Instead, they simply used machine learning as a tool in a new environment, which is insufficient for patent eligibility. The court rejected Recentive's argument that applying machine learning to a new field (such as event scheduling or network mapping) rendered the claims non-abstract, reiterating that limiting an abstract idea to a particular field of use does not make it patent-eligible. 

Step Two: Inventive Concept?

After finding that the claims were directed to a “patent-ineligible concept,” the Court moved to the next step of the Alice analysis: determining whether the claims posses an “inventive concept” that is “sufficient to ensure that the patent in practice amounts to significantly more than a patent upon the ineligible concept itself.”  The Court found that the claims here did not. 

According to the Federal Circuit, the use of machine learning to dynamically generate and update schedules or maps was, in the court’s view, simply the abstract idea itself. The claims did not describe any new or improved machine learning techniques, nor did they specify how the machine learning models achieved their results beyond generic, functional language. In particular, according to the Court, the claims did not disclose “a specific implementation of a solution to a problem in the software arts” (quoting Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1339 (Fed. Cir. 2016)) or “a specific means or method that solves a problem in an existing technological process” (quoting Koninklijke KPN N.V. v. Gemalto M2m GmbH, 942 F.3d 1143, 1150 (Fed. Cir. 2019)). 

The Court also rejected the argument that increased speed and efficiency from computer implementation could supply an inventive concept, as this is inherent in using computers for any task. 

Key Takeaway for Patenting Machine Learning Inventions

This decision provides a clear and important message for companies and inventors seeking to patent machine learning innovations:

Patents that merely apply generic machine learning techniques to new data environments or fields of use—without disclosing specific improvements to the machine learning models or methods themselves—are not patent-eligible under § 101. 

To be eligible, a machine learning patent must claim a specific technological improvement, such as a novel machine learning algorithm, architecture, or training method—not just the use of existing machine learning tools to automate or optimize tasks previously performed by humans. The court’s holding underscores the need for patent applicants in the machine learning space to focus on concrete technical innovations in their claims and specifications, rather than broad functional applications of known techniques. 

This decision aligns with prior Federal Circuit precedent that abstract ideas do not become patent-eligible simply by being implemented on a computer or applied to a new field. Companies developing machine learning solutions should carefully consider the technical substance of their inventions and ensure that their patent applications clearly articulate specific, non-generic improvements to machine learning technology.