AI and Patent Eligibility: Strategies in the Wake of Recentive Analytics v. Fox Corp.
Legal Alerts
4.23.25
The rapid growth of AI and Machine Learning (ML) models across industries has raised important questions about the scope of patent eligibility under 35 U.S.C. § 101. Recently, in Recentive Analytics, Inc. v. Fox Corp.,[1] the Federal Circuit addressed this issue, evaluating whether patents that apply ML to optimize a technology meet the threshold for patent-eligible subject matter. The Federal Circuit held that the patents were directed to abstract ideas and lacked any inventive concept that would transform them into patent-eligible subject matter. This article examines the court’s reasoning and outlines practical takeaways for AI-related patent strategy.
Key Takeaways
- Generic application of ML to new data contexts is insufficient: Merely applying known ML models to different environments or fields does not make a claim patent-eligible.
- Technological improvement is essential: Patents must disclose how the invention improves the functioning of the computer or the ML technique itself, not just achieve desirable results.
- Efficiency gains are not enough: Increased speed and efficiency from using ML compared to human processes does not itself create eligibility.
- Avoid purely functional claiming: Functional claim language describing desired outcomes (e.g., “generate optimized schedules”), without detailing how to achieve those outcomes in a non-abstract way, is insufficient for patentability.
- Concrete implementation matters: Patents should clearly explain how any purported improvements to ML technology are accomplished, not just state that improvements exist.
- Describe intended solution or improvement: Claims should describe a specific implementation of a solution to a problem in the software arts or a specific means or method that solves a problem in an existing technological process.
Patent eligibility under 35 U.S.C. § 101 requires that an invention fall within one of four statutory categories: process, machine, manufacture, or composition of matter. Even if this threshold is met, the invention must not be directed to a judicial exception (e.g., a law of nature, natural phenomenon, or abstract idea), unless it includes an “inventive concept” that adds significantly more to transform it into a patent-eligible application. This analysis is guided by the Supreme Court’s two-step Alice/Mayo framework: first, determining whether the claims are directed to a patent-ineligible concept; and second, evaluating whether the claims contain an “inventive concept” sufficient to transform the abstract idea into a patent-eligible application, adding “significantly more” than the judicial exception itself.[2]
In Recentive Analytics v. Fox Corp., Recentive, the plaintiff-appellant, asserted four patents (U.S. Patent Nos. 10,911,811, 10,958,957, 11,386,367, and 11,537,960) covering the use of ML models to generate optimized network maps and event schedules for television broadcasts and live events. These patents not only covered the use of ML models but also the training of these models. The Machine Learning Training patents (‘367 and ‘960 patents) focus on the process of training ML models, including iterative training steps to identify relationships in event data and improve the model’s accuracy over time. The Network Map patents (‘811 and ‘957 patents) primarily claim the application of a trained ML model to generate and optimize network maps. Recentive argued that these patents were eligible under § 101 because they applied ML techniques in a novel way to a specific domain, producing dynamically generated and updated outputs. Fox Corp., one of the defendants-appellees, countered that the patents merely applied conventional ML methods to a new data environment without offering any improvement to the underlying technology. The district court agreed and granted Fox’s motion to dismiss.
Applying the Alice framework, a Delaware District Court first found that the asserted claims were directed to the abstract ideas of creating network maps and event schedules using generic mathematical techniques. At step two, it concluded that the claims lacked an inventive concept, relying instead on broadly described, conventional ML methods implemented on generic computing hardware. The court emphasized that automating an existing manual process with standard ML tools does not amount to a technological improvement. Unlike the claims in McRO v. Bandai,[3] which recited specific, non-conventional rules that enhanced computer animation, the Recentive patents failed to describe any similarly concrete innovation. Accordingly, the court held that applying known ML techniques to a new field (e.g., TV scheduling) without advancing the ML technology itself was insufficient under § 101. It also denied Recentive’s request to amend its complaint, finding amendment would be futile given the ineligibility of the asserted claims.
Recentive appealed the decision to the Federal Circuit. In its opinion affirming the district court’s decision, the court implicitly contrasted two distinct claiming strategies since some of the asserted claims were directed to the actual training of ML models, albeit using conventional techniques, while others merely recited the use of AI to generate outputs like schedules or network maps. The court’s treatment of both underscores an important point: even claims involving the training of AI are not automatically eligible if the techniques used are routine or described in abstract terms. On the other hand, simply invoking AI as a tool to achieve a result, without detailing how the AI is applied in a novel or technically meaningful way, is also insufficient. Thus, whether claims are directed to training processes or the application of AI outputs, eligibility hinges on disclosing concrete, technical improvements and not just broad functional goals or the repackaging of routine methods.
Notably, the court also observed that while the claims disclosed the use of ML in a new environment (event scheduling and the creation of network maps), they failed to describe “a specific implementation of a solution to a problem in the software arts,”[4] or “a specific means or method that solves a problem in an existing technological process.”[5] These are both strategies that courts have recognized as ways to satisfy the eligibility requirements of § 101.
Ultimately, the court confirmed that to survive § 101 scrutiny, AI-related patents must do more than apply existing ML techniques to new problems; they must disclose how the technology itself is improved. This includes not only improvements to the operation of computer systems or software generally, but also enhancements to the underlying ML models or training processes. This aligns with the broader framework established by the courts, which recognizes that claims directed to a judicial exception may nonetheless be patent-eligible if they demonstrate a practical application that improves the functioning of a computer, ML system, or other technological process. In contrast to claims that merely recite functional outcomes or automation of existing tasks, patent applicants should clearly articulate the specific, technical solutions their inventions provide. Going forward, those seeking to protect AI innovations should ensure their claims are tied to concrete implementations that address technological problems, not merely the application of algorithms to new data environments.
[1] Recentive Analytics, Inc. v. Fox Corp., 1:22cv1545.
[2] Alice Corp. Pty. v. CLS Bankint ‘l, 573 U.S. 208 (2014); Mayo Collab. Servs. v. Prometheus Lab’ys, Inc., 566 U.S. 66 (2012).
[3] McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299 (Fed. Cir. 2016).
[4] Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1339 (Fed. Cir. 2016).
[5] Koninklijke KPN N.V. v. Gemalto M2M GmbH, 942 F.3d 1143 (Fed. Cir. 2019).