Earlier this month, the Federal Circuit heard oral argument in Thaler v. Vidal, a case we have been following because it attracted an amicus brief. Thaler is the developer, user, and owner of DABUS, an artificial intelligence system that created the two inventions at issue in the case. On appeal, Thaler seeks review of a district court’s grant of summary judgment to the U.S. Patent and Trademark Office, which decided that an artificial intelligence machine cannot be an “inventor” under the Patent Act. This is our argument recap.
Ryan Abbott argued for Thaler. Abbott argued that the Patent Act uses terms that can refer to natural persons, but not terms that must refer to natural persons, including the term “individual” used to define “inventor.” Additionally, he cited § 103 of the patent statute, which indicates “patentability shall not be negated by the manner in which the invention was made.”
Abbott claimed that an “individual” can be anything, including a company. Chief Judge Moore, however, countered this claim by pointing out that companies cannot file patents. In response, Abbott argued that “corporate inventorship is different than AI inventorship” because, unlike recognizing corporations as inventors, with an AI invention there is no risk of excluding a human inventor that contributed to the invention.
Judge Taranto then pressed Abbott on the dictionary definition of artificial intelligence, saying that “AI” refers to the intelligence of a machine and not the machine itself. Abbott said that he uses the term “AI” to describe a “distinct inventive entity,” one software program run on one specific computer.
Judges Stark sought to return to the language of the patent statute by asking if there are any indications that “individual” does not mean “just natural persons.” Abbott argued that case law supports the idea that context and purpose are critical for statutory interpretation. He also said that denying patents in areas not contemplated by Congress would frustrate the purposes of patent law.
Judge Taranto asked how to determine whether broadening the definition of inventor to include pieces of software would advance or frustrate innovation. In response, Abbott provided an example of Pfizer using a super-computer to make an antibody to treat COVID. He said in this example the “inventive activity” would be the activity of the computer, and if a patent would not be allowed, this would send a message to companies that they cannot use computers in research and development if they care about obtaining patents. But, he argued, the patent statute was “designed to incentivize innovation” and “promote disclosure of trade secrets and commercialization of new products,” and a narrow interpretation of “inventor” would frustrate these purposes.
Dennis C. Barghaan Jr. argued for the USPTO. He began by pointing out that only a decade ago Congress amended the patent statute to “explicitly define ‘inventor’ as an individual.” The amendment, he argued, “expressly bolstered that definition by using personal pronouns to define that individual.” In response, Judge Taranto emphasized that the amendment uses gendered personal pronouns.
Barghaan then cited precedent defining “individual” as a human being and refusing to expand that definition. He explained that the case analyzed the term using its dictionary definition, its normal usage, and the Dictionary Act. Based on this case, Barghaan argued that, without written explicit language that says “individual” should be interpreted broadly, the term should apply only to natural persons.
Barghaan then went on to refute Abbott’s point regarding § 103, saying that the provision only refers to the idea of conception, not to the patent statute as a whole. The purpose of the section, he argued, is to avoid preclusion of patentability based on use of routine experimentation, not to make the whole patent statute flexible.
Judge Taranto then asked whether a scientist using AI to develop an invention may file a patent application in his or her own name. Barghaan answered that, under the facts of this case, Thaler had no involvement whatsoever in the conception of the invention. As a result, Barghaan, maintained, Thaler cannot claim himself as the inventor. Barghaan conceded that if Thaler could say he participated in some way in creating this subject matter and listed himself as the inventor, that might be a different story.
Chief Judge Moore asked if the programmer of AI code should be the inventor, not the program itself. She offered an example involving gene sequencing. Barghaan responded that a programmer of AI code should be the inventor in limited circumstances including this example.
In rebuttal, Abbott first addressed the question of whether the programmer of AI might qualify as an inventor in some circumstances, and he agreed with Barghaan that it is possible. He, however, warned about complications of allowing programmers to qualify as inventors given that there are situations where the original programmer may have nothing to do with subsequent inventions created with the AI. When asked by Judge Stark about § 103, Abbott agreed that the provision was enacted to deal with the “flash of genius problem,” but, he argued, § 103 is broader than Barghaan claimed. Abbott concluded his rebuttal by arguing that rewarding innovation should be the focus, whether that innovation comes from “a person or a robot or a room full of monkeys,” although he immediately backtracked to exclude monkeys.
We’ll continue to monitor the case and report on any developments.