Chase Webber, MJLST Staffer
In the last year, a United States Patent and Trademark Office (USPTO) and a US federal district court opinion answered many questions on the issue of who (or what) should get the patent when AI is responsible for an invention? The take-away from these new sources is that AI cannot be named as the “inventor” for a patent, but the human developers of the AI will qualify as joint inventors. Future courts (including the appeal of Thaler v. Hirshfeld, the recent AI “inventor” case) may consider whether an invention created by an AI may be practically unpatentable because no “natural person” could qualify as the inventor. However, according to the USPTO report, humans’ significant contributions to the AI in all existing instances of AI invention qualify them as the “joint inventors” to obtain a patent. There is no need to change the patent laws today for AI. It is essential to consider, however, which humans will qualify as inventors.
Stephen Thaler applied for a patent on behalf of his AI, DABUS, for two inventions generated by DABUS. The inventions are inconceivable by human minds; for example, a beverage container designed using fractal geometry. Thaler named DABUS the “sole inventor” when filing for a patent. On DABUS’s behalf, Thaler assigned all DABUS’s patent rights and benefits to himself. The USPTO refused to issue the patent to DABUS because an “inventor” must be a “natural person.” Facing an insurmountable obstacle of basic statutory interpretation, Thaler argues policy: The USPTO’s interpretation disincentivizes innovation and devalues human inventorship – both of which threaten the purpose of the Patent Act. Thaler’s arguments do not persuade the court or this author.
Who really cares? The outcome seems to be the same for Thaler – whether he is assigned DABUS’s patent rights or awarded the patent for DABUS’s invention. The public commentary in the USPTO report, published before the case and referenced by the court in Thaler, seems to agree. There is no apparent injustice in a human taking credit for the AI’s invention. However, the outcome of Thaler is important when considering who is responsible for (and therefore profits from) AI inventions.
Thaler claims that the AI is the rightful “inventor,” although, practically, someone besides the AI will profit no matter the outcome of Thaler. But who? Is it the owner of the AI machine? Or the developers who created the AI (assuming these are separate people)? Both the AI owners and developers? According to the USPTO report, it’s more likely to be the developer, but it may be both. Joint inventorship is determined by considering whether a person’s contribution to the invention was significantin the scope of the entire invention. AI developers easily qualify for joint inventorship. However, AI owners may only claim joint invention if they can show that their contribution to the AI (e.g., the introduction of crucial data which the AI used in its invention process) was significant to the AI’s invention. These issues are determined on a case-by-case basis. Thaler, for example, is the owner and developer of DABUS, therefore, Thaler would be the “sole inventor” on the patent application for DABUS’s inventions.
Future plaintiffs in AI “inventor” cases should frame their case around joint ownership, instead of sole ownership. In this way, the court would consider which humans could profit from AI inventions instead of AI personhood, a theory doomed by philosophical reasoning if not only textualism. Using the “significant contribution” standard, the AI owner can never legally be named “sole inventor” in exclusion of the AI developers as joint inventors. In contrast, naming the AI itself as “sole inventor” (as Thaler proposes) allows the AI owner to assign patent rights to himself and exclude the AI developers from benefitting.
A court could be more generous if it compares the implication of Thaler’s proposal to the Patent Act goals in the context of joint inventorship. Using the USPTO’s interpretation, if AI machines are marketed to numerous owners, the developers profit from patents earned by each individual machine, whether or not any individual owner will profit. Under Thaler’s interpretation, the one who owns the AI and applies for a patent will benefit from the AI’s invention. In that case, the developers will only profit from the initial sale of the machine, not from the patent of any AI inventions. The implication of the USPTO’s interpretation may threaten innovation, as a plaintiff of Thaler’s position might argue, because the individual owners have little incentive to produce valuable and innovative AI-generated inventions if they likely cannot profit from the patents of those inventions.
What happens if no humans can show that they made significant contributions to the AI? According to public comments summarized by the USPTO, there is no such legal circumstance now. To some, there never will be. When the Patent Act says, “Whoever invents . . . may obtain a patent . . .” (emphasis added), it refers to he who contributes significantly to an AI invention (explained above). AI is legally only a tool used by a natural person, who is himself the inventor. The non-AI computer is a familiar example of a tool that can be used for invention, where it is obvious that the computer itself is not the “inventor.” Thaler seems to imply that since no human could conceive DABUS’s inventions (e.g., the fractal geometry), DABUS is no longer a tool but the inventor himself. Philosophically – and, as the court determines by statutory interpretation, legally – it seems clear that DABUS is only a tool used by Thaler, especially when considering that Thaler (and not DABUS, as he concedes) stands to profit.
AI (including DABUS) is limited to “narrow, application-specific objectives” in contrast to artificial general intelligence (AGI) which resembles human intelligence. To picture AGI, imagine the AI portrayed in sci-fi movies, e.g., Ex Machina, AI that can pass The Turing Test. Could AGI hypothetically graduate from “tool” to “inventor”? The public opinion gathered by the USPTO is split on this question: One side argues “no.” Even AGI, they claim, has the same practical human origins as AI; there is no legal difference in patent law. The other position is “who cares?” To argue whether AGI is an “inventor” is to claim AGI’s moral personhood, a question best left for the field of philosophy that far exceeds the scope of this discussion. One day, we may confront this issue in legal or technological reality. For now, AGI is still a matter of imagination and cinema.