Recentive Analytics, Inc. v. Fox Corp.

Recentive Analytics v. Fox Corp., decided April 18th, 2025. A panel of the Federal Circuit, in an opinion by Judge Dyk, affirmed the dismissal of four machine-learning patents as ineligible for patenting. The court held that claims that do no more than apply generic machine learning to a new data environment, without improving the machine learning itself, are directed to an abstract idea and are not patent eligible. Here's the brief.

Recentive Analytics owns four patents that use machine learning in the entertainment and broadcast industries. Two of them — the "Machine Learning Training" patents — generate optimized schedules for live events. The other two — the "Network Map" patents — generate network maps, which determine the programs a broadcaster's channels display in particular markets at particular times.

The claimed methods share a common structure. A machine-learning model is trained on historical data, receives user-defined inputs and target features, generates an optimized schedule or map, and updates that output in response to real-time changes. The patents themselves say any suitable, conventional machine-learning technique will do.

Recentive sued Fox Corporation and two affiliates for infringement in the District of Delaware. Fox moved to dismiss, arguing the patents claimed ineligible subject matter under Section 101. The district court agreed, dismissed the case, and denied leave to amend as futile. Recentive appealed to the Federal Circuit.

Section 101 defines what may be patented, and the Supreme Court has long read it to exclude abstract ideas.

Courts apply the two-step Alice test. First, ask whether a claim is directed to an abstract idea. If it is, ask second whether the claim adds an inventive concept that transforms it into something significantly more. The court framed this as a question it had not decided before.

At Alice step one, the court held the claims directed to an abstract idea. Recentive conceded it was not claiming machine learning itself, and the patents described only conventional, generic machine-learning technology run on generic computers. Iterative training and dynamic updating, the court said, are incident to the very nature of machine learning — not a technological improvement. And Recentive admitted its patents did not claim any method for making the machine learning itself better.

Applying settled precedent, the court held that limiting an abstract idea to a new field of use, or to a novel database, does not confer eligibility. Neither the claims nor the specifications described how any improvement to the machine learning was achieved; they recited only the result. That, the court said, distinguished the patents from cases like McRO, where the claims disclosed a specific technique different from the one humans used. Nor does performing a task humans once did with greater speed and efficiency make a claim eligible.

At step two, the court found no inventive concept. Recentive's asserted concept — using machine learning to dynamically generate optimized maps and schedules — was no more than the abstract idea itself. The court affirmed the dismissal and the denial of leave to amend.

Recentive is the Federal Circuit's first decision on the patent eligibility of machine-learning claims. It holds that applying generic, off-the-shelf machine learning to a new kind of data does not make an abstract idea patentable; the claims must disclose an improvement to the machine-learning models or techniques themselves. The court was careful to limit its holding.

Recentive Analytics v. Fox Corp., 134 F.4th 1205, decided April 18th, 2025. I'm Kyle Coleman. Thanks for watching.