Must an “inventor” be a human being? Indeed, in Thaler v. Commissioner of Patents the U.S. District Court for the Eastern District of Virginia upheld the view of the United States Patent and Trademark Office: Artificial intelligence (AI) algorithms cannot be inventors for the purposes of a U.S. patent. Through a textualist lens, the court turned to numerous sources, including Congress’ definition of “inventor” in the Patent Act and 2011 America Invents Act (AIA) as an “individual,” multiple dictionaries, and common usage, to ultimately conclude that “individual” referred to a “natural person.” Thus, an inventor would need to be a natural person.

Importantly, however, the decision concludes that AI may “someday reach a level of sophistication sufficient to establish inventorship,” a task for Congress, not the courts. Furthermore, Stephen Thaler (plaintiff) announced he intends to appeal the decision before the Federal Circuit. This captivating case may in fact compel an expansion of the term “inventor,” and only time will tell if AI will have its day as Named Inventors in the US.

Elsewhere, however, courts have concluded that AI systems are eligible for inventorship. A pioneer in the movement, South Africa has granted a patent for an AI-created invention, and Australia follows its path, both rulings challenging the US decisions. Nevertheless, concluding that AI systems qualify as inventors is a long shot. When seen in close-up, South Africa handles patents entirely different than the US; it automatically grants all patent application that are filed without meaningful substantive examination. On the other hand, zooming in on the case in Australia reveals a broad interpretation of inventor as an “agent.” Further, an agent was defined as “any person or thing that invents.” Accordingly, while the applicant is the owner of the AI technology, if the AI system is the agent, it can be an “inventor.”

By expanding the definition of “inventor” to promote technological innovation and the publication and dissemination of such innovation by rewarding it, one of the defining purposes of patent protection is fulfilled. However, if patent law fails to lend a hand to AI systems, then the intellectual property landscape is in for a non-natural earthquake, as owners may seek to protect patentable inventions of this sort as trade secrets, especially given that copyright law requires the author to be a human being with moral rights. As expected, a defining feature of trade secrets is that a trade secret can be any business or technical information that has economic value to the user, is not generally known by or available to the public, and is subject to reasonable efforts to preserve its secrecy. Thus, such technology would be shielded from public disclosure.

In fact, we would have trouble fitting certain AI, such as training data sets used to inform a machine-learning model, into intellectual property doctrines. Indeed, the only potential option for data sets of this kind would be protection through trade secrets.

Without question, much uncertainty comes with AI technology. Accordingly, it is increasingly important for regulation in this high-tech galaxy to protect innovation.

I assert a novel regime of intellectual property to protect facets of AI… a regime sans prior use.