AI systems are increasingly capable of creating works and inventions without the aid of a human inventor. In July 2019, the Surrey team garnered worldwide attention when it filed two patent applications at several patent offices, including the USPTO, naming an AI system, the Device for the Autonomous Bootstrapping of Unified Sentience ("DABUS"), as the inventor. DABUS is a creativity machine that has the ability to generate inventions. It can also determine which inventions are most useful. According to the patent applications, DABUS invented—without human assistance—an improved beverage container and a "neural flame" device used in search-and-rescue missions.
The DABUS applications raised broader questions about whether IP generated by undirected AI should be entitled to legal protections. In August 2019, the USPTO issued a notice seeking public comment on the patenting of AI inventions. Shortly thereafter, it issued a second notice seeking similar comment on the availability of legal protections for other forms of AI-created IP. Among other things, the USPTO sought input on how to ensure that appropriate incentives were in place to encourage further innovation in the realm of AI-related technologies.
In the meantime, the Office began processing the Surrey team's applications. After an initial review, the USPTO determined that the applications were incomplete because they failed to identify a human inventor. In response, the Surrey team submitted petitions seeking supervisory review. After the Office dismissed those petitions in December 2019, the Surrey team moved for reconsideration, contending that inventorship should not be limited to natural persons.
The Office ultimately denied the motions for reconsideration of the petitions in the two applications, and published one of those decisions in late April. The USPTO emphasized that the relevant statutory provisions governing patents consistently refer to inventors as natural persons. The Office reasoned that a plain reading of those provisions precluded a finding that the term "inventor" can refer to an AI system. The USPTO also cited several Federal Circuit precedents stating that only natural persons can be inventors. In University of Utah v. Max-Planck-Gesellschaft zur Forderung der Wissenschaften E.V., 734 F.3d 1315, 1323 (Fed. Cir. 2013), for example, the Federal Circuit rejected a claim that corporations or sovereigns could be inventors. Because the relevant statutes and case law, as well as USPTO rules and regulations, limit inventorship to natural persons, the Office concluded that the Surrey team's patent applications could not list an AI system as the inventor.
As part of its decision, the USPTO also acknowledged several public policy arguments offered by the Surrey team. The team maintained that allowing an AI system to be listed as an inventor would (i) incentivize innovation; (ii) reduce the improper naming of inventors; and (iii) promote public notice of an invention's actual creator. These considerations, however, were insufficient to overcome the plain meaning of U.S. patent laws as construed by the courts.
While the USPTO's decision provides clarity that an AI system cannot be an inventor under current law, it leaves many questions unanswered. For example, it remains unclear whether and to what extent legal protections are available to owners of inventions conceived with the assistance of AI. Likewise, as the decision arose in the context of patents, it obviously does not address other forms of IP created without human aid. More broadly, the decision meaningfully adds to the ongoing dialogue about the implications of AI on longstanding concepts of IP rights under U.S. law while raising fundamental questions as to whether existing laws need to be revamped to address new and emerging technologies. Stakeholders should monitor changes in this sphere and watch for additional guidance from the USPTO.