by Kurt Buechle

Artificial Intelligence (AI) is revolutionizing a broad array of technologies, not just by automating routine tasks, organizing and inputting data, and performing programmed algorithms, but by solving problems and performing jobs in a manner approximating those performed using intelligent thought. And in so doing, AI is creating new ways and approaches for solving those problems and performing those jobs.  Although AI is generally carried out in a virtual environment on a computer, the structures and processes employed in AI often find analogies in and/or are inspired by components and operations of electrical, mechanical, chemical, and biological systems—the use of neural networks by AI being just one example.

In addition to posing a broad range of technical and ethical concerns, AI has raised many questions that directly impact the United States Patent and Trademark Office’s consideration of whether or not to grant AI-related patents.  Such questions include, for example, how AI is defined, whether AI constitutes patent eligible subject matter, and whether AI itself can qualify as an inventor.  The United States Patent and Trademark Office (USPTO) is addressing these questions on several fronts: broadly engaging the public through requests for comment, specifically responding to particular patent applicants during prosecution, and by creating an AI-dedicated webpage. This webpage provides information on the USPTO’s AI initiatives, public notices and responses, decisions, AI-related events, and outside resources.

In its Request for Comments on Patenting Artificial Intelligence (Federal Register, Vol. 84, No. 166; August 27, 2019), the USPTO sought comments (during a comment period since closed) on patenting AI inventions, including both inventions that utilize AI and inventions developed using AI.   The USPTO asked whether patent laws and regulations need to be revised if an inventor is an AI rather than a natural person.  The USPTO also asked about issues relating to how AI inventions can comply with various requirements for obtaining a patent, for example, subject matter eligibility, written description, and enablement, as well as how the ordinary skill in the art should be evaluated.

Various legal organizations as well as corporations in a broad range of industries responded to the USPTO’s Request, framing concerns regarding the consistent treatment of AI throughout the international patent community (the responsive Comments can be found on the USPTO’s AI webpage).  Several areas of concern highlighted in the Comments include ambiguity of certain terms, both as they are used in the U.S. and internationally, subject matter eligibility, and incentivizing use of AI.  For example, the Intellectual Property Organization (IPO) addressed the ambiguity that the term “AI invention” poses as it can refer to an invention that utilizes AI or to an invention that is created by AI.  To address this ambiguity, the IPO proposes to use the term “Core AI invention” to refer to technologies that have general applicability to the field of AI, the term “Application Specific AI invention” for inventions that use Core AI technologies to address a specific problem or to perform a particular kind of task, and the term “AI Generated invention” for those inventions actually created by AI.  (Comments on Patenting Artificial Intelligence Inventions, November 11, 2019.)  The first two terms and their respective technological categories mirror definitions provided by the Japan Patent Office (JPO) in their comments.  The JPO has already released guidance on the examination of AI inventions.  (JPO Comments to Patenting Artificial Intelligence Inventions, September, 2019.)  Defining an AI invention is complicated by the fact that AI itself does not have an agreed upon definition as pointed out by IBM in its comments (IBM Corporation Comments, November 8, 2019).

The European Patent Office (EPO) questioned the subject matter eligibility of AI given an understanding that AI is based on computational models and algorithms that have “an abstract mathematical nature.”  But, the EPO acknowledged that AI inventions could be patentable if they have a technical character as determined by how the computations models and algorithms are implemented.  (Comments by the EPO.)  IBM in its comments also voiced concern about the subject matter eligibility of AI inventions.  (IBM Corporation Comments, November 8, 2019.)  The JPO does not believe that there are any patent eligibility considerations unique to AI inventions. The JPO makes the assumption that the USPTO’s 2019 Revised Patent Subject Matter Eligibility Guidance (Guidance) will allow for suitable protection of AI inventions. (JPO Comments to Patenting Artificial Intelligence Inventions, September, 2019.)  The Guidance was issued in January 2019, updated in October 2019, and will likely see further revisions, some of which may address AI directly.  As has been the case with computer-implemented inventions generally, how claims are drafted and the art unit in which an application is examined may impact whether a particular AI invention is viewed as subject matter eligible by the Office.

Companies that often apply for patent protection for their innovations look to the USPTO to ensure that patent policies for innovation continue to grow and adapt to afford protection to technological advances including those in AI. For example, Novartis explained in its comments that AI is transforming the way the company innovates and operates.   Novartis stated that the patent system must be able to both incentivize AI inventions and inventions developed using, and potentially by, AI.  Novartis argues these principles are grounded in the “Constitutional objective of promoting scientific and technological process.”  A focus on such fundamentals is seen as a bulwark against the limitations of existing rules and policies stifling AI inventions.   (Comments by Novartis, November 8, 2019).  In recognizing the potential for AI to change the inventive landscape, potential patentees also recognized that AI advances may also change how inventive entities are viewed.  IBM questioned whether AI could ever be an inventor because a computer does not have the capability of recognizing the significance, value, or both of an invention.  (IBM Corporation Comments, November 8, 2019).  The USPTO recently addressed whether AI can be an inventor, not based on AI’s capability to recognize an invention, but rather based on the requirement that an inventor must be a natural person.

On April 27, 2020, the USPTO posted a petition decision (“Decision”) that directly addressed the issue of AI as a named inventor in a patent application.  In the Decision, the USPTO held that, under current law, only natural persons can be named as an inventor in a U.S. patent application.  The petitioned application listed an artificial intelligence as the sole inventor along with a substitute statement in lieu of a signed inventor’s declaration.  The USPTO had found that the application did not name the inventor(s) by his/her legal name.  The applicant, also the purported assignee and the signatory on the substitute statement, petitioned that finding, eventually leading to the Decision.  In its opinion for the Decision, the USPTO relied on 35 U.S.C. § 115 as requiring the naming of the inventor and asserted that Title 35 consistently refers to inventors as natural persons.  The USPTO further cited Federal Circuit decisions holding that natural persons cannot be corporations or sovereigns, and extended that reasoning to exclude machines as natural persons.  Among other arguments of the petitioner rejected by the USPTO, the USPTO was not persuaded that the artificial intelligence could or should be an inventor because patents claiming that particular artificial intelligence had been issued.

Views on AI inventions will undoubtedly evolve as AI itself evolves and raises new legal questions and other issues.  We look forward to following future developments and learning more about how the USPTO will approach AI as these developments occur.