Inventors filing patent applications relating to artificial intelligence ("AI") and machine learning ("ML") technologies now have greater clarity of the requirements for obtaining patent protection.
On April 18, 2025, the U.S. Court of Appeals for the Federal Circuit (CAFC) issued its opinion in Recentive Analytics, Inc. v. Fox Corp. et al., the court's first opinion on patent eligibility for AI and ML technologies. The CAFC's decision highlights the importance of patent applicants' filing thorough and complete disclosures for their AI and ML inventions.
Recentive Analytics, Inc. had brought a lawsuit against Fox Corp., Fox Broadcasting Company, LLC, and Fox Sports Productions, LLC, claiming infringement of four of Recentive's patents. Two of the Recentive patents are directed to the training of ML technologies to schedule live events, while the other two Recentive patents relate to generating broadcast network maps with ML technologies. The CAFC affirmed a finding by the United States District Court for the District of Delaware that the Recentive patents were directed to patent ineligible subject matter.
In the opinion, the CAFC walked through the two-step Alice inquiry used by courts to determine patent eligibility and found the Recentive patents wanting at both steps. In its analysis under step one of the Alice inquiry, the CAFC determined that the Recentive patents merely claimed an abstract idea, namely the application of existing AI and ML technologies to a process or machine that already substantially existed. The court noted that the Recentive patents failed to claim or disclose any feature that improved the AI, machine, or process claimed. According to the CAFC, the Recentive patents' requirements that its ML model "be 'iteratively trained' or dynamically adjusted" were not improvements, but instead were aspects "incident to the very nature of" AI. The CAFC also noted that the Recentive patents did not claim any "specific method for 'improving the mathematical algorithm or making machine learning better.'" Additionally, there was nothing in the claims or specifications of the Recentive patents explaining how any improvement to the claimed AI processes and systems was accomplished.
The CAFC concluded that the Recentive patent claims amounted to a use of AI in a "new environment," and such new uses of existing technologies are not patent eligible. Further, the court stated that improvements in speed, efficiency, or accuracy over the prior art by themselves are insufficient for the AI invention to be patent eligible. Under step two of the Alice inquiry, the CAFC stated that it could "perceive nothing in the claims . . . that would transform the [Recentive patents] into something 'significantly more' than the abstract idea of generating event schedules and network maps through the application of machine learning."
While the CAFC said much about the ways Recentive failed to render its claimed AI inventions patent eligible, the court explicitly noted that there may yet be AI inventions that "lead to patent-eligible improvements in technology." The level of disclosure required for AI inventions to be patent eligible may be discerned from the CAFC's statements of what the Recentive patents lacked, specifically that the level of disclosure is generally a high bar to clear. Any claim to an AI invention will need enough disclosure to show that the AI technology is functioning in a way that is different from the operation of conventional AI systems. Examples of necessary disclosure may include outlining the AI platform being used, the specific process by which the AI learns or trains, the inputs for the AI, the types of data used to see the AI, and the process by which the AI solves a problem or reaches a solution. Additional details such as the specific improvements that the AI invention enables compared to current technologies may be needed to show that the AI invention does more than merely allow for otherwise known processes or machines to work faster or more efficiently. The disclosure of the specific improvements is even more important if the invention relies on a conventional AI or ML process (i.e., defining input data for the AI, training the AI on the input data, and updating/improving the AI over time based on the input) that is being implemented in novel ways.
In sum, a detailed disclosure is needed for any patent application claiming the use of AI. The disclosure needs to be specific and provide significant details such as the AI platform used, the inputs for the AI, the kind of data is used to seed the AI, the process by which the AI reaches a solution, the method of training or learning for the AI, and/or the improvement on the prior art enabled by the use of AI (other than speed, efficiency, or accuracy).
The Pryor Cashman team will continue to monitor the latest changes in patent law to ensure our patent prosecution and litigation practice provides the highest quality of service. Should you have any questions regarding patent eligibility or other patent law-related concerns, please contact one of the attorneys listed below.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
