Mikele Bicolli
Oct 28, 2022

CAFC Confirms Again That Artificial Intelligence Cannot Be an Inventor | When Will AI Break Patent Law?

Created at Oct 28, 2022, 9:54am PDT


In 2019, Stephen Thaler filed two patent applications, 16/524,350 (teaching a “Neural Flame”) and 16/524,532 (teaching a “Fractal Container”). What was unusual in these applications was that he named an AI system called DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the sole inventor. He maintains that he didn't contribute to these inventions as he merely took the output from DABUS and reduced it to practice. To satisfy the mandatory requirements of the patent application Thaler submitted a statement of DABUS as a sworn oath or declaration by the inventor. The UPSTO, however, concluded that both applications were incomplete. As a result, they sent two “Notices to File Missing Parts of Nonprovisional Application,” concluding that no valid inventor was provided in the application. Thaler submitted a petition to USPTO to vacate these notices, but the petition was denied, and UPSTO maintained that the machine does not qualify as an inventor.

The matter evolved in district court litigation between the parties. After hearing the oral arguments, the court decided to favor the defendants and against Stephen Thaler. It was concluded that under the United States Patent Act, an inventor must be an individual and, according to the court,  a human being. The decision was appealed further to the Court of Appeals for the Federal Circuit in September 2021. At that time, it looked like this case could either end up being decided very quickly with a short reading of the Patent Code or result in a major litigation and important decision with significant implications for the future of AI and innovation. On August 5, 2022, the district court’s decision was affirmed by the CAFC. In its ruling, the Federal Circuit focuses on interpreting the terms “inventor” and “individual.” The Patent Act defines the inventor (or inventors) as an individual (or individuals) who discovers the invention. Although the Patent Act doesn't provide an explanation for the word “individual'', the Federal Circuit cites a 2012 Supreme Court decision that determined that an individual is considered a human being. Thaler had included in the complaint that South Africa had granted patents with DABUS as an inventor. However, this argument was shut down given that South Africa doesn't interpret the U.S. Patent Law, so its determination couldn't alter the CAFC conclusion. In this way, it was decided that only a human being can be considered an inventor, not AI. It’s unclear if the case will be brought to the Supreme Court since the CAFC mentioned a prior decision supporting their conclusion, although the case being litigated was not related to deciding whether AI can be an inventor in the U.S.

This decision, for now, ranks the U.S in the list of countries that don’t allow AI to be an inventor. Thaler has submitted applications to the European Patent Office, Germany, the UK, Australia as well as South Africa, which as mentioned above, granted the patents. EPO rejected and decided that the application must contain “a family name, given names and full address of the inventor who must be a human being.” The United Kingdom Court of Appeal also decided against Thaler, ruling that “Thaler did not comply with the requirements of the patent application,” which required him to identify a person as the inventor and to indicate how he had derived his rights from that person. Australia, meanwhile, by interpreting its Patent Law, didn't find any obstacle to not allowing a non-human inventor. Given that non-human entities cannot file a patent application, the applications by Thaler were viewed as correct and consistent with Australian Patent Law by the Federal Court of Australia, which reverted the initial decision of the Commissioner of Patents to deny the patent applications.

When will the AI “break” Patent Law? This is a difficult question to answer, but in the near future, this issue needs to be addressed as AI further develops, and we will see more of these applications naming a non-human entity as the inventor. For now, a solution was suggested by the German Federal Patent Court to name the inventor as the person behind the process and the AI as a tool used to come up with the invention. This could be the case in Stephen Thaler's applications, given that the inventions were patentable and the rejection was based only on the fact that the inventor couldn't be the machine. The other direction would be to change the patent and intellectual property laws in a way that they would accept non-humans as inventors and authors. However, this will raise the issues of ownership of the patent rights or royalties that come with AI patents or creations. As of now, AI is considered as software or hardware, and a person or organization has ownership over them. In this way, everything produced by them is owned by their owner. It’s unclear how the ownership concept will work if we decide to raise the AI rights to the human level by allowing them to have patents. Is it moral to own an entity that is able to create, invent and make profits?