The United States Patent & Trademark Office (USPTO) launched its Artificial Intelligence (AI) Initiative in 2019. In furtherance of the Initiative, the USPTO solicited public comments on patent-related AI issues. The USPTO published the comments shortly before the COVID-19 pandemic. We anticipate additional activity in this area as the government and economy emerge from restrictions, and the industry once again fully focuses on research and development. Below, is an analysis of major themes arising out of the public comment process.
The USPTO requested feedback on various issues impacting AI and the patent laws, ranging from general issues of patent protection to inventorship to disclosure issues. The USPTO received comments from 43 organizations and 55 individuals.
Several overarching themes emerged from the responses even if none of the questions posed by the USPTO garnered unanimous agreement. At the highest levels, new AI technologies and technologies incorporating AI should be eligible for intellectual property (IP) protections. The incentive system of the U.S. IP regime rewards innovation and investment in development of new and useful technologies. In this sense, AI is no different from many transformative technologies that came before AI.
Accordingly, there was general consensus that non-human entities—such as AI machines—should not be afforded inventorship status nor allowed to “own” patent rights. The most common arguments include an AI machine’s lack of ability to conceive, inability to perform inventive steps independently, and the notion that AI is akin to a process or tool used to create a new product or result. Underlying these arguments is the notion that if AI machines were allowed to be named as inventors and/or owners of patented technologies, it is possible that the entities funding that development would be denied the benefits of those investments. As Telefonaktiebolaget LM Ericsson commented, “[t]he party that paid for the AI, made the AI, or paid for the related context in which the AI was utilized to make the invention should be able to own any resulting patents.” Traditional employment and IP transfer agreements between natural persons and entities funding and supporting research properly distribute the various patent rights from any discoveries. Allowing AI machines to qualify as inventors or owners of IP would turn this system on its head. Shortly following publication of the public comments, the USPTO ruled that inventorship is limited to natural persons in an application naming an AI machine as the inventor. (Appl. Ser. No. 16/524,350 (the “DABUS” application), Petition Dec., April 22, 2020).
Further, most comments suggested that no major changes to the patent laws are necessary to address issues raised by AI technologies. One group of commenters noted that the patent laws have accounted for major technological shifts in the past, and there’s no reason to think that AI will be any different. These commentators likened AI inventions to software and computer inventions, commenting that similar eligibility impediments may arise. Indeed, IBM commented, “[j]ust as computers are inherently abstractions of a tool, AI is also an abstraction; an abstraction of the human brain, human learning, human thought. . . . [T]he patent eligibility of AI inventions is at risk under the current abstract idea doctrine.” Another group of commenters took the position that AI technologies are still developing and that it’s too early—and perhaps entirely unnecessary—to modify existing statutory and regulatory regimes. For example, the Association of University Technology Managers, Inc. (AUTM) captured that sentiment by stating, “[c]urrent patent examination framework (with specific guidance) is workable as applied to AI-enabled inventions, at least for now.”
Submissions varied with respect to the USPTO’s more detailed questions regarding AI’s impact on patent disclosure requirements. For instance, a majority of organizational submissions did not see unique enablement or written description issues arising from AI technologies, but were more closely split on the questions of prior art and POSITA considerations. The submission by American Intellectual Property Lawyers Association (AIPLA) encapsulated the issue:
If it is decided that inventive AI may be an inventor, this decision may have a profound impact on the determination of the level of ordinary skill in the art for purposes of a validity analysis. The factual analysis of determining the ordinary skill may require an additional assessment of the state of the art of AI at the time of invention. . . . If it is determined that only natural persons may be inventors, the existing analysis is workable in that it can run on the assumption that those working in the art use the relevant available tools, whether these are AI or other tools.
While the comments addressed various additional nuances and issues, in general, the innovation community does not see a need for statutory or regulatory changes to the patent law to address AI issues…yet.
In sum, the public comments responding to the USPTO show that we are still in the infancy (or at least the toddler phase) of AI development and that no need exists for statutory or regulatory change…yet. The USPTO has not taken any further actions on the comments or on its AI Initiative—likely due to the slowdown mandated by the COVID-19 crisis. We will continue to monitor the USPTO’s response and look forward to the evolution of U.S. IP laws as AI technology continues to expand.