Facts
Recentive Analytics owns four US patents on the use of machine learning in television broadcasting. Two, referred to as the “Machine Learning Training” patents, claim generating optimised schedules for live events by iteratively training a model on historical data and updating the schedules in real time. The other two, the “Network Map” patents, claim using machine learning to generate and update network maps that determine which programmes a broadcaster's stations show in particular markets at particular times. Recentive sued Fox for infringement in the District of Delaware. The district court dismissed the case under 35 U.S.C. § 101 and refused leave to amend.
Question
The question was whether claims applying established machine learning techniques to a new field are directed to patent-eligible subject matter under the two-step test in Alice v. CLS Bank. At step one, the court had to decide whether the claims are directed to an abstract idea, and at step two whether they contain an inventive concept that transforms that idea into a patent-eligible application. Recentive argued that iterative training, dynamic adjustment to real-time data and the application of machine learning to event scheduling and network maps made its claims eligible.
Decision
Writing for the panel, Judge Dyk affirmed. Recentive had conceded that it was not claiming machine learning itself, and the patents described conventional techniques running on generic computers. Iterative training and dynamic adjustment are incident to the very nature of machine learning and are not technological improvements. Applying machine learning to a new field of use or data environment does not make an abstract idea non-abstract, and performing a task once done by humans faster or more efficiently does not create eligibility. At step two, the asserted inventive concept was no more than the abstract idea itself. The refusal of leave to amend was upheld because Recentive had proposed no amendment that would change the analysis.
Why it matters
The court framed its holding narrowly: patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the models, are ineligible under § 101. It added that machine learning is a burgeoning field that may lead to patent-eligible improvements in technology. The court stressed that the claims did not delineate steps through which the machine learning technology achieves an improvement; claims that describe only the desired result do not meet that standard. The Supreme Court declined to review the case.
Related stages
Panel: Judges Dyk and Prost, with Chief District Judge Goldberg sitting by designation. District court: D. Del. No. 1:22-cv-01545-GBW, 692 F. Supp. 3d 438 (2023). The Federal Circuit denied rehearing on 23 July 2025, and the Supreme Court denied certiorari on 8 December 2025 (No. 25-505). For a UK ruling that took a different approach to an artificial neural network invention, see emotional-perception-uk-supreme-court; on whether an AI system can be an inventor, see thaler-dabus-uk-supreme-court and dabus-japan-ip-high-court.