USPTO Releases Public Comments on AI

November 11, 2020 – Article

The USPTO recently released a report on public comments that were received in response to two requests for comments (RFCs) regarding artificial intelligence (“Public Views on Artificial Intelligence and Intellectual Property Policy”). The USPTO states that it will use this report to focus issues for continued exploration of other measures it may take to bolster the understanding and reliability of IP rights for emerging technologies, such as AI.

As background, the U.S. Patent and Trademark Office (USPTO) has sought to engage the public in the emerging field of Artificial Intelligence in the past couple of years. In January 2019, the USPTO held the AIIP Policy Conference which included panel discussions on patents, trade secrets, copyrights, trademarks, IP enforcement, global perspectives, and the economics of IP protection of AI. Following the conference, the USPTO proactively solicited public input through two formal Requests for Comments (RFC) published in the Federal Register. In response, the agency received approximately 200 distinct comments from a broad range of experts in foreign patent offices, bar associations, trade associations, academia, law firms, and companies in the electronics, software, automobile, medical, and pharmaceutical industries.

Specifically, the first RFC, issued on August 27, 2019, requested comments on patenting AI inventions and the USPTO received 99 comments in response. The second RFC, issued on October 30, 2019, requested comments on other (non-patent) IP policy areas such as copyrights, trademarks, database protections, and trade secret law. The USPTO received approximately 98 comments in response to the second RFC. 

The first RFC sought to understand the elements of an AI invention. The comments from the public were summarized in the report as defining AI as “computer functionality that mimics cognitive functions associated with the human mind (e.g., the ability to learn).” More importantly, the report summarized AI inventions as following into the following three categories:

(a) inventions that embody an advance in the field of AI (e.g., a new neural network structure of an improved machine learning (ML) model or algorithm)

(b) inventions that apply AI (to a field other than AI),

(c) inventions that may be produced by AI itself.

While a discussion of AI elements and categories of AI inventions is usually part of a background discussion on this topic, the USPTO has the power to put these answers to consequential use in the form of examiner training and art unit classification. Classification can in itself be the ball game, especially when one thinks of the two paths of patent prosecution that hinge on being classified as a business method patent or not. 

The always hot topic of inventorship of AI was addressed, and the majority of comments suggested that current AI could neither invent nor author without human intervention. The comments suggested that on the technical nature of AI innovations, human beings remain integral to the operation of AI, and this is an important consideration in evaluating whether IP law needs modification in view of the current state of AI technology. The comments appeared to be consistent with the positions taken by the USPTO earlier this year in the DABUS case that an inventor must be a natural person. However, some comments suggested that a machine could still conceive of an invention, albeit AI technology may not be sufficiently advanced at this time to exclude the role of a human inventor.

On subject matter eligibility, the comments reflect what many patent practitioners have been emphasizing the past several years. Specifically, the comments suggest that AI inventions should not be treated differently than other computer-implemented inventions, but AI inventions are at a higher risk under a subject matter eligibility analysis because they related to human activity, mental processes, or mathematical concepts.   

An interesting patent theme was raised in the report which stressed that it may be difficult to enable (i.e., teach the public to make and use) certain AI inventions, as required by 35 U.S.C. § 112(a). Regarding the written description requirement, some commenters indicated that there are significant challenges to satisfying the disclosure requirements for an AI invention because even though the input and output may be known by the inventor, the logic in between (such as a neural network) is in some respects unknown. While the report is brief on this issue (relative to other topics in the report), this may become a significant issue based on the practical goals of claiming strategy. For example, for an AI invention directed to object recognition, the claim may have a scope of coverage for what objects can be recognized that is broader than the range of objects that can actually be recognized by the AI invention described in the specification. This will be a challenging judgment call for Examiner’s to make, and we will provide a blog post in the future which goes expands upon this issue in greater detail.

Regarding non-patent issues, most commenters acknowledged that existing copyright law does not permit a non-human to be an author. Some commenters noted that AI is a tool, similar to Photoshop, than can be used to create works. It appears that only a minority of commenters suggested that a sufficiently creative work made by AI without human intervention should be copyrightable with authorship being taken by either in the owner/controller of the AI system or in the person/user who fixes the work in its final form.

Turning to another copyright topic, “Ingestion” is a process where an AI algorithm learns its functions by “ingesting” large volumes of copyrighted material. Commenters believe that this does raise an infringement issue, but the existing law is adaptable to this issue on a case-by-case basis with infringement or fair use both being a possibility.

Across all IP topics, a majority of public commenters expressed a general sense that the existing U.S. intellectual property laws are calibrated correctly to address the evolution of AI. However, commenters may diverge on whether any new classes of IP rights would be beneficial to ensure a more robust IP system, and what the implications of more advanced AI may bring. While there is no legal significance to the report, the report provides an excellent primer on the important IP issues facing AI innovations.