Skip links
Patent AI Inventions

How to Patent AI Inventions

Patent Office Rules Support AI Invention Patents

Artificial Intelligence (AI) has revolutionized numerous industries, offering innovative solutions and enhancing efficiencies. However, with these advancements comes the challenge of encouraging innovators to develop AI technology including to patent AI inventions. Promoting responsible innovation, competition, and collaboration in AI will allow the United States to lead in AI development and unlock the technology’s potential to solve some of society’s most difficult challenges. This effort requires investments in AI-related education, training, development, research, and capacity, while simultaneously tackling novel intellectual property (IP) questions and other problems to protect inventors and creators.

Recognizing that “[r]esponsible AI use has the potential to help solve urgent challenges while making our world more prosperous, productive, innovative, and secure,” while “[a]t the same time, irresponsible use could exacerbate societal harms such as fraud, discrimination, bias, and disinformation; displace and disempower workers; stifle competition; and pose risks to national security,” President Biden issued the “Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” on October 30, 2023 (Executive Order). To help ensure innovation in the area of AI continues and in response to Pres. Biden’s Executive Order 14110, the U.S. Patent and Trademark Office (USPTO) recently issued guidance addressing this critical issue, providing much-needed clarity for inventors and businesses alike in patenting their AI Inventions. The USPTO guidance addresses the patent elligibility for patents on AI inventions. The following is a discussion about analyzing AI inventions for patent elligiblity based on the recent patent guidance.

Understanding the Abstract Idea Exception

In the U.S. and around the world, patent protection is applicable to a process, machine, manufacture, composition of matter, combination or improvement of the same. Generally, patent protection does not extend to abstract ideas, laws of nature or natural phenomena. No one owns the exclusive right to ideas, thoughts and formulas and thus they are not patentable. Generally, they include concepts or thoughts which are not tangible and can include mathematical concepts, certain methods of organizing human activity (e.g., fundamental economic principles; commercial or legal interactions; managing personal behavior, relationships or interactions between people) and mental processes.

With respect to mental processes and Patenting AI Inventions, the Patent Office provides the following clarification: “[C]laims do not recite a mental process when they contain limitations that cannot practically be performed in the human mind, for instance when the human mind is not equipped to perform the claim limitations. … [C]laim limitations that only encompass AI in a way that cannot practically be performed in the human mind do not fall within this grouping.”

The abstract idea exception is a judicially created doctrine that excludes certain subject from patent protection. However, under the Supreme Court’s opinion in Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014), in some cases, inventions which include an abstract idea can still qualify for patent protection, if they do something more. The abstract idea exception can pose a significant hurdle for AI inventions, as many involve algorithms and data processing techniques which could be considered abstract.

Alice/Mayo Framework

Under Alice the Supreme Court established a framework for analyzing whether inventions in various technologies, including Computer and AI inventions, should be considered patentable. This framework applies to all categories of claims, i.e.. process, machine, manufacture and composition of matter claims and all types of judicial exceptions, including, abstract idea, laws of nature and natural phenomena. The Alice/Mayo test is broken down into a number of steps which are illustrated in the following flow charts.

patent AI inventions

Step 1 of the Alice/Mayo analysis address whether the claimed invention falls into one of the four main categories of patentable inventions; processes, machines, articles of manufacture or compositions of matter. Step 2 looks at whether the invention, even though covered in one of the above categories, falls into one of the judicial exceptions for patentable inventions; i.e. is the invention directed to a law of nature, natural phenomena or abstract idea under Step 2A. If the answer is no, the invention is directed to patent elligible subject matter.

Patent AI Inventions

If the answer is yes, then the Alice/Mayo analysis looks at whether the invention amounts to more than a judicial exception. If the invention uses the judicial exception as part of a practical application of the judical exception, the analysis continues to the second prong of Step 2A, illustrated to the left.

The analysis of whether the invention integrates the judicial exception into something patentable in Step 2A includes factors such as whether the additional elements are a significant extra-solution activity; the claim reflects an improvement in the function of a computer or another technology or technical field.

By integrating the judicial exception into a practical application of the judicial exception, the invention is no longer a judicial exception but is directed towards patentable subject matter. The analysis continues under Step 2B to look at whether the invention provides an inventive step or is lacking, i.e. insignificant, well-understood, routine or conventional. If the invention lacks an inventive step, it fails under step 2B. Some of the practicle applications (from various court decisions) include:

  • Whether the invention provides insignificant extra-solution activity?
  • Whether the invention is merely an instruction to apply the abstract idea?
  • Whether the invention reflects an improvement in the functioning of a computer?
  • Whether the invention reflects an improvement to another technology or technical field?
  • Whether the invention provides a technical solution to a technical problem?
  • Whether the invention merely improves the abstract idea (e.g., is the claim merely directed to an improved mathematical process)?

If the invention does not integrate the abstract idea into a practical application of the abstract idea or lacks an inventive step, the invention is not patent eligible subject matter.

The analysis under Step 2A, Prong Two and Step 2B are similiar, except Step 2B considers whether the additional element (or combination of elements) is a well understood, routine or conventional activity which lacks patent coverage. The Alice/Mayo test for determining if an invention includes patentable subject matter does not change when considering an AI invention.

Patents for AI Inventions

The Patent Office has provided guidance as it relates to three different types of AI patents, one for a neural network, one for speech filtering and a third one related to an AI model designed for medical treatment. Based on the guidance, each invention includes some claims directed toward patent elligible inventions and some claims directed to patent inelligible inventions. The following is just a summary, but the detailed analysis from the patent office goes through each invention for hypothetical claims, on a limitation-by-limitation basis and for the claim as a whole.

The neural network invention is generally directed to an invention using AI to detect data anomalies from expected data or patterned data. Claim 1, directed to an integrated circuit that included a plurality of neurons organized in an array and a plurality of interconnecting synaptic circuits. Claim 1 was found to be patent elligible under Step 2A, Prong One because it was not directed to an abstract idea under Step 2A. Claim 2 was directed to an artificial neural network that included steps of receiving data, seperating the training data, tranning the neural network, detecting anomalies in the network and outputting the result. Claim 2 was not patentable subject matter becasue it was an example of mental processes. However, Claim 3 which was also directed to the neural network was patent elligible under Step 2A, Prong Two because it was more specific and integrated the abstract idea of the neural network and provided a practicle application into something which improved the function of a computer in the technical field of detecting anomolies in networks.

The speech filtering invention was generally directed to an invention which used AI to analyze speech signals and seperate desired speech from background speech. Claim 1 was directed to a method of speech seperation which includes the steps of receiving speech, converting speech and using a deep neural network to embed vectors from mixed speech signals. Claim 1 was patent inelligible because it merely combined an abstract idea and did not do anything more. Claim 2 provided details on how the deep neural network provided for clustering by separating different speech sources and synthesizing speech from these clusters and was found to be patent elligible because it did something more under Step 2A, Prong Two. Claim 3 is directed to a storage medium which causes a processor to preform the steps of receiving a mixed speech signal, convert the mixed speech signal, cluster and mask the speech source, convert the masked clustered speech to filter the speech and extract the background noise from the speech to produce a transcript of the speech. Claim 3 was found to be patent elligible because it integrated the abstract idea into a practicle application which was not insignificant.

The AI medical model invention was generally directed to a Compound X, a treatment for a glaucoma patients. Claim 1 was directed to a treatment method which was found to be patent inelligible becasue it was simply a combination of abstract ideas and improvements to those ideas. There was no technological improvement involved in Claim 1. However, Claim 2 was found to be patent elligible because it used Compound X for treatment which as a whole converted the abstract idea into a practile application under Step 2A, Prong Two.

Conclusion

As inventors try to Patent AI inventions, many patent applicants will need an experienced patent attorney to address the issues related to subject matter eligibility. If you or your company are seeking patents on AI-related inventions, contact us. Our firm is ready to help you navigate these murky waters. In addition to helping our clients obtain strong patents, our familiarity with Patent Office policy also enables us to provide wholistic advice to clients facing patent disputes, including defending against AI-related patents.