The Federal Circuit's recent decision in Recentive Analytics, Inc. v. Fox Corp. has established new boundaries for patent eligibility in machine learning applications, potentially reshaping how companies protect their artificial intelligence innovations. The April 18, 2025 ruling affirmed a lower court's dismissal of patent infringement claims, holding that applying existing machine learning models to new data environments without disclosing improvements to the underlying technology fails to meet patent eligibility standards under 35 U.S.C. § 101.
Court Rejects Generic Machine Learning Applications
The case centered on four patents held by Recentive Analytics related to using machine learning models for generating television broadcast schedules and network maps. Fox Corp. successfully argued that these patents were ineligible under Section 101, with both the District Court of Delaware and the Federal Circuit agreeing that the claims represented abstract subject matter under the Alice framework.
The Federal Circuit emphasized that "claims that do no more than apply established methods of machine learning to a new data environment, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101." The court found that Recentive's patents relied on generic, conventional machine learning models and training techniques without providing any improvement to the underlying ML technology itself.
Crucially, the court determined that simply training a machine learning model is "incident to the very nature of machine learning" and does not represent a technological improvement. The ruling stressed that even if machine learning applications improve speed or efficiency compared to manual human processes, this automation alone does not render claims patent eligible.
Technical Innovation Requirements Clarified
The decision provides important guidance for patent applicants in the AI space. The court made clear that invoking a computer to automate processes traditionally carried out by humans does not satisfy patent eligibility requirements, even when such automation delivers performance improvements. In Recentive's case, the tasks of generating schedules and network maps had been performed manually by humans using similar data types before machine learning implementation.
The Federal Circuit's analysis highlighted several critical factors that undermined Recentive's patent claims. The company conceded it was "not claiming machine learning itself" and that the patents' specifications would be satisfied by "any suitable machine learning technique." Most significantly, Recentive acknowledged that the patents did not disclose "a technological improvement" on the machine learning techniques applied to event scheduling and network mapping tasks.
Implications for Patent Strategy
The ruling raises important questions about what constitutes sufficient technological improvement in machine learning patents. While the court suggested that "improvements to machine learning models" can support eligibility, key uncertainties remain regarding the threshold for such improvements. Patent practitioners must now consider whether demonstrating enhanced accuracy, efficiency, or scalability constitutes adequate technical innovation, or whether courts will view such benefits as merely mathematical in nature.
The decision also leaves open questions about the magnitude of improvement required for patent eligibility and how much weight courts will give to empirical evidence showing improved model performance. As courts continue applying the Alice framework rigorously, applicants must ensure their patent applications clearly articulate technical innovations beyond new use cases for existing machine learning tools.
Trade Secrets Emerge as Alternative Protection
The Recentive decision may accelerate the shift toward trade secret protection for AI innovations. Both the Defend Trade Secrets Act and state law analogues offer broader protection for machine learning applications, covering information that provides economic value through secrecy and is subject to reasonable protection efforts.
Trade secret protection can extend to various aspects of machine learning systems, including methods for configuring models to specific use cases, training data, model weights, and outputs. Unlike patents, trade secrets are not limited to human-created information and do not require upfront disclosure or filing fees.
However, companies pursuing trade secret protection face the challenge of precisely defining their protected information. Trade secret owners must identify secrets with reasonable particularity and implement adequate protections through employee training, nondisclosure agreements, confidentiality markings, and information security measures.
Strategic Considerations for AI Companies
The Federal Circuit's decision underscores the need for AI companies to carefully evaluate their intellectual property strategies. Patent applications must now demonstrate clear technical improvements to machine learning models themselves, rather than simply claiming novel applications of existing techniques.
For companies developing machine learning applications, the ruling suggests focusing patent efforts on innovations that advance the underlying technology, such as new training algorithms, novel model architectures, or improved optimization techniques. Meanwhile, trade secret protection may be more appropriate for proprietary implementations, training datasets, and application-specific configurations that provide competitive advantages through secrecy.
The decision also highlights the importance of engaging intellectual property counsel early in the development process to establish robust frameworks for identifying and protecting valuable AI innovations, whether through patents or trade secrets.
