On July 17, 2024, the U.S. Patent and Trademark Office (PTO) issued additional guidance focusing on the subject matter eligibility of Artificial Intelligence (AI) inventions. This updated guidance (the “update”) was issued in response to President Biden’s Executive Order 14110 recognizing the need to foster responsible innovation in AI, while “tackling novel intellectual property (IP) questions and other problems to protect inventors and creators.” The USPTO’s stated goal for the guidance update “is providing a guidance to promote clarity, consistency, and address innovation in AI and critical and emerging technologies.”
The update notes that the existing procedures apply to all technologies, including AI, and the update covers the overall subject matter eligibility analysis, including specific instructions for evaluating AI inventions under the well-known Alice/Mayo test. The guidance update addresses AI by focusing on Prongs One and Two of Step 2A of the test.
Prong One includes determining whether a claim recites an abstract idea. Here, the update notes that it is common for claims reciting AI inventions to involve abstract ideas (e.g., mathematical concepts, certain methods of organizing human activity, and mental processes). However, the update cautions that “USPTO personnel must draw a distinction between a claim that ‘recites’ an abstract idea (and thus requires further eligibility analysis) and one that merely involves, or is based on, an abstract idea.” For example, as set forth in the examples included in the update, claims to an application specific integrated circuit for an artificial neural network may not recite an abstract idea and may not require further eligibility analysis. In contrast, software-based implementations of an artificial neural network may recite an abstract idea and, thus, may require the full eligibility analysis. The update also notes that “[t]he mental processes grouping is not without limits, and as such, claim limitations that only encompass AI in a way that cannot practically be performed in the human mind do not fall within this grouping.”
Prong Two evaluates whether the claim integrates the judicial exception into a practical application. Previous guidance on this prong indicated that one way to demonstrate such a practical application was to show that the claimed invention improves the functioning of a computer or another technology or technical field. The update states that “[m]any claims to AI inventions are eligible as improvements to the functioning of a computer or improvements to another technology or technical field.” However, as with existing computer implemented inventions, this improvement must be shown beyond “instructions to implement the judicial exception on a computer” or the general linking of the judicial exception to a particular technological environment or field. In other words, it is important to show that the claims provide a particular solution to a problem and not the idea of a solution.
The update also introduces three new examples (47–49) to illustrate the application of the subject matter eligibility analysis to three sets of AI related claims. The examples include various AI technologies. Example 47 relates to an artificial neural network for anomaly detection, Example 48 relates to deep learning for signal processing, and Example 49 relates to a machine learning model for personalizing medical treatment. Each of the examples features eligible and ineligible claims.
In the Examples, claims are eligible when they (i) recite specific hardware implementations, (ii) apply abstract ideas using particular techniques that improve a technology, or (iii) involve particular implementations of machine learning models in practical applications that have “meaningful limits.” The Examples illustrate how including specific technical details about the structure, training, or implementation of AI models can transform an otherwise ineligible claim into an eligible one by demonstrating a practical application that improves technology.
Based on the update, applicants should consider the examples when drafting new applications and during prosecution when facing abstract idea rejections. For example, aligning an overall claim structure to one of the examples may help an eligibility argument. Furthermore, the update clearly demonstrates the importance of describing, in the specification, how the claimed invention improves existing technology. Detailed specifications are also helpful to support arguments that a claimed AI invention cannot practically be performed in a human mind and also to show that the claims provide a particular solution to a problem and not just the idea of a solution. For example, describing an AI-based invention at a high level (e.g., without details of the specific model used, how the model is trained, and the like) may undermine later arguments that the invention provides a particular solution to a problem and may limit potential claim amendments to not only overcome cited art but also overcome an abstract idea rejection.
It is important to note that the update does not have the force of law, and the courts continue to wrestle with this issue. However, inventors seeking protection for AI-based inventions may at a minimum appreciate that the USPTO considers AI inventions to be critical technologies and will continue to tackle questions and challenges for protecting such inventions.