IP Law Daily, SUPREME COURT NEWS—Supreme Court keeps intact machine learning patent ruling, (Dec 9, 2025)

By George Basharis, J.D.

The Court declined review of Recentive Analytica’s petition challenging the Federal Circuit’s conclusion that applying existing machine learning methods to new applications does not meet the threshold for patent protection.

The U.S. Supreme Court’s refusal to review a challenge to the Federal Circuit’s machine learning patent ruling leaves in place an interpretation of Section 101 that limits patent protection for software inventions applying established learning models to new industries. The denial maintains the appellate court’s conclusion that Recentive’s patents for dynamically generating network maps and live event schedules claim abstract ideas and fail to recite an inventive concept. The outcome preserves a precedential decision that treats the use of known machine learning techniques in new environments as insufficient for patent eligibility, an approach that Recentive argued diverged from the preemption principles underlying the Supreme Court’s Alice Corporation v. CLS Bank International, 573 U.S. 208 (2014), decision. The Court denied Recentive’s petition without comment on December 8, 2025.

Background. Recentive sued Fox Corp., Fox Broadcasting Company, LLC, and Fox Sports Productions, LLC in the District of Delaware for allegedly infringing four patents directed to methods for dynamically generating television network maps and live event schedules by applying iteratively trained machine learning models to complex, data-rich scheduling problems. The patents fall into two groups: the Network Map patents, which describe processes for assigning events to stations across cities, and the Machine Learning Training patents, which outline iterative training techniques for creating optimized live event schedules that respond to real-time changes.

The district court dismissed the complaint, concluding that both patent families were directed to abstract ideas. In the court’s view, the claims involved producing network maps and schedules through mathematical techniques implemented on generic computer systems and did not include the type of inventive concept required at Alice step two. The court found that the asserted improvements were rooted in the use of known machine learning algorithms rather than in any technological advance to the models themselves.

The Federal Circuit affirmed, holding that the claims did no more than apply conventional learning methods to a new data environment and therefore fell on the ineligible side of Section 101. The court of appeals framed the question as one of first impression and concluded that limiting a known technique to live event scheduling or broadcasting did not supply eligibility. It also agreed that Recentive’s arguments at step two recited the abstract idea itself rather than an inventive concept. Recentive’s petitions for rehearing and rehearing en banc were denied.

Issues raised in the certiorari petition. Recentive’s petition for certiorari asked the Supreme Court to review two questions: whether the Federal Circuit’s eligibility analysis conflicts with the Supreme Court’s instruction to evaluate preemption concerns, and whether the appellate court erred in treating applications of machine learning to new data environments as categorically ineligible unless the claims recite improvements to the underlying models.

The petition asserted that the Federal Circuit’s decision deviates from the statutory breadth of Section 101 and from Alice’s emphasis on avoiding monopolization of the basic tools of scientific work. According to Recentive, the Federal Circuit’s analysis allowed the eligibility exceptions to expand beyond their limited purpose. The company argued that its claims teach concrete, domain-specific processes for weighting, training, and dynamically updating models in response to evolving scheduling constraints. It further contended that the claims target technical barriers that existed not because of a lack of computational power but because prior systems could not operate dynamically across millions of potential map configurations.

The petition highlighted the scheduling challenges that arise in industries such as professional sports broadcasting, where hundreds of live events require coordinated mapping across markets and time slots. It described the claimed technology as incorporating iterative training, user-specific weighting, and model updates that react to real-time changes in inputs. Recentive argued that these claimed steps produce demonstrably more accurate outputs than prior manual or static mapping methods and therefore constitute patent-eligible improvements.

Recentive also warned that the Federal Circuit’s approach poses particular risks for machine learning and artificial intelligence advances. The petition stated that many real-world innovations arise not from revising underlying neural architecture but from adapting known models to new data, constraints, and domains. It contended that the decision below improperly converts the use of general-purpose processors and off-the-shelf learning tools into a basis for ineligibility, which the petition characterized as contrary to the Patent Act and the Supreme Court’s precedent.

Although the petition framed the case as an opportunity to restore preemption principles to the center of the eligibility analysis and to provide guidance for AI-driven innovation, the Court declined to take up the issues.

News: Patent FedCirNews SupremeCtNews