The patent landscape for artificial intelligence (AI) is on the cusp of major shifts. Recent developments, including the U.S. Patent and Trademark Office’s (USPTO) 2024 Guidance Update on Patent Subject Matter Eligibility, specifically addressing AI, and the potential passage of the Patent Eligibility Restoration Act of 2023 (PERA), could significantly reshape the field.
What is PERA?
The Patent Eligibility Restoration Act of 2023 aims to overhaul the patent eligibility framework under U.S. patent law by clarifying and simplifying the criteria for what qualifies as patentable subject matter. The Act seeks to simplify and standardize patent eligibility determinations; resolve widespread confusion among courts, agencies, and patent practitioners; and provide a clearer and more predictable framework for innovators.
The key objectives include:
Elimination of Judicial Exceptions: The Act eliminates the judicially created exceptions (e.g., abstract ideas, laws of nature, and natural phenomena) that have historically narrowed patent eligibility, leading to confusion and inconsistency.
Restoration of Broad Patent Eligibility: Any invention that falls into one of the statutory categories—process, machine, manufacture, or composition of matter—and has practical utility is patentable, subject to limited exclusions.
The following are not eligible for patents under PERA:
Mathematical formulas unless integrated into a larger, practical invention.
Mental processes that occur solely in the human mind.
Unmodified human genes as they exist naturally.
Unmodified natural materials as they exist in nature.
Purely economic, financial, business, social, cultural, or artistic processes unless they inherently require a machine or physical manufacture to operate.
PERA further clarifies on the following:
Modified Natural Elements Can Be Patentable: Isolated, purified, enriched, or otherwise altered genes or natural materials, as well as their application in a useful invention, may qualify for patent protection.
Whole-Invention Analysis: Patent eligibility is assessed by considering the invention as a whole, without disregarding any claim element regardless of how was the invention made, whether the claim element is known, conventional, routine, or naturally occuring, the state of the art, or any other consideration in section 102, 103, or 112.
Patent Infringement Cases: Courts can determine patent eligibility at any stage of an infringement case, with limited discovery allowed to resolve questions of eligibility.
In November 2024, Senators Thom Tillis (R-NC) and Chris Coons (D-DE) announced that they would delay consideration of PERA and encouraged opponents to engage constructively in the process in the meantime.
If PERA becomes law, it would expand the range of inventions eligible for patent protection, especially in fields like biotechnology and artificial intelligence. This change could lead to a surge in AI-related patent filings.
The Challenge of Subject Matter Eligibility
Subject matter eligibility is a foundational requirement for patent protection. To qualify, AI inventions must demonstrate meaningful innovation beyond simply analyzing data to produce a model. Many AI inventions are categorized as abstract ideas under judicial exceptions, leading to stricter scrutiny compared to inventions in other fields where written descriptions and prior art are the primary hurdles.
Given the current landscape, innovators should take proactive steps to file patent applications promptly. Early filings can secure enterprise value and ensure operational freedom as laws evolve. By adopting strategic approaches to meet patent eligibility criteria, applicants can maximize their chances of success.
Strategies for Achieving Patent Eligibility
Currently, a claimed invention must meet the requirements of one of four statutory categories (process, machine, manufactured item, or composition of matter) and satisfy one of three standards to achieve subject matter eligibility:
No Judicial Exception: The invention is not directed to judicial exceptions such as laws of nature, natural phenomena, or abstract ideas.
Practical Application: If an invention involves a judicial exception, it can still qualify if additional elements integrate the exception into a practical application.
Inventive Concept (aka "Significant More"): If neither of the above applies, the invention can qualify by demonstrating significantly more than the judicial exception.
For AI technologies, the key question often centers on whether the invention recites an abstract idea. If PERA passes, the elimination of this judicial exception could simplify the path to eligibility.
Strategies for AI Patent Applications
Frame the Invention as a Technical Solution: Position the invention as a technical solution to a specific problem, particularly one that improves computer functionality or addresses challenges in specific technical fields. For example:
Avoid describing the invention by its desired result; focus instead on what problem it was designed to solve and how it achieves that result.
Detail data structures, algorithms, and technical improvements, such as reduced processing times or enhanced security.
Explicitly connect claim limitations to hardware components, such as specialized processors, to bolster patent eligibility.
However, routine steps and structures would not amount to a technical improvement, particularly when described in the specification as merely automating previously manual tasks. See Miller Mendel, Inc. v. City of Anna, Texas, No. 2022-1753 (Fed. Cir. July 18, 2024).
Target Favorable USPTO Art Units: The assigned USPTO art unit can influence the success of a patent application. For example, applications tied to medical informatics, for example, may perform better in the biotechnology-focused 1600 Technology Center than in the e-commerce-focused 3600 Technology Center, which has higher rejection rates.
The field, claim preamble, and other sections of a patent application specification can help position an invention within a favorable art unit and strengthen eligibility arguments under the practical application or inventive concept aspects of the analysis.
Carefully draft the specification and claims to guide the invention’s classification into a preferred art unit. Avoid terminology that might lead to unfavorable classifications.
Detail AI Model Training Processes: Examiners often scrutinize the specifics of AI training. Applications that include well-defined, iterative training processes with detailed data inputs are more likely to meet eligibility requirements.
The USPTO’s 2019 Revised Patent Subject Matter Eligibility Guidance offers valuable insights for drafting eligible claims, as demonstrated in Example 39. This example avoids the classification of a mental process as an abstract idea since the described steps cannot be practically carried out in the human mind. The claim focuses on an iterative training algorithm, where the system undergoes a second training stage using an updated dataset that includes false positives identified during the initial training stage.
Avoid Conventional Computer Functions and Mathematical Formulas: Claims that focus solely on data collection or storage are likely to be deemed abstract. Therefore, obtaining data should be recited passively or excluded. Data storage limitation should also be avoided when possible.
Refrain from explicitly claiming routine computing functions or mathematical formulas unless integrated into a practical application.
The USPTO confirmed in Example 48 of the 2024 guidelines that comparisons and correlations arguably fall outside of the mathematical concept abstract idea.
If mathematical concepts must be included, ensure they are integrated in a practical application as in Example 45 of the USPTO’s 2019 guidelines.
Leverage Practical Uses for AI Outputs: Instead of merely describing the generation or storage of AI model outputs, focus on how the outputs are used to solve a technical problem.
Claims that apply outputs to practical and technical applications, even for routine tasks, are more likely to qualify for patent protection. For instance, specifying a particular use for an AI-generated prediction can enhance eligibility.
Example 49 from the 2024 guidelines provides valuable insight. Claim 1, which states “administering an appropriate treatment,” is deemed ineligible for patent protection. However, Claim 2, which specifies “the appropriate treatment is Compound X eye drops,” qualifies as patent-eligible. Although administering eye drops is a well-known and routine activity, this fact is irrelevant to the practical application prong of the patent eligibility analysis. This distinction underscores that merely stating a desired outcome without detailing the specific means to achieve it may render a claim ineligible.
When drafting AI-related patent claims, it’s crucial to move beyond generic descriptions of functions or results. For instance, instead of claiming “an AI system that identifies objects,” one should specify the particular methodologies employed, such as “an AI system utilizing a convolutional neural network trained on a dataset of labeled images to identify objects.” This approach ensures that the claim is rooted in concrete technical implementations rather than abstract ideas.
Looking Ahead
With potential legislative changes on the horizon and evolving USPTO guidance, the AI patent landscape remains dynamic. Strategic preparation and adherence to patent eligibility best practices can help applicants navigate uncertainties and secure robust IP protections in this rapidly advancing field.
Comments