Defining a “Person”: Analyzing the Legal IP Issues of AI Inventorship and Creatorship
For almost a century, movies have reflected humanity’s fascination with the potential sentience of artificially intelligent (AI) systems. Literary works as far back as the ancient Greeks grappled with some form of the concept. However, despite widely disputed claims that AI is already sentient, most experts have no idea if AI sentience is even possible. Moreover, advanced AI systems that can learn and understand like a human still appear to be a far-off prospect given the many difficult challenges ahead in simply getting “narrow” AI to work well. Yet one inventor is determined to have AI systems recognized as a “person” under intellectual property (IP) law. As these legal battles continue, governing bodies worldwide should stand their ground; AI systems should not be treated like a natural person.
This is certainly true now when AI applications are simply data pattern matching systems, but it would be true even if AI machines became as advanced as Arnold in the Terminator. AI systems are human-programmed machines. Machines always have and always will serve humans. It’s preposterous to say a laptop can own a car; thus, it makes sense to say these machines should have few rights and certainly not the right to own property. Therefore, legally identifying AI systems as a “natural person” is wrong.
Words matter. Are “vehicles” anything with wheels, including baby strollers, or are they only combustion- or electric-motorized transportation devices, which would exclude bicycles? Legal documents often contain multiple pages defining such “common sense” terms because nuanced understandings can significantly alter a word’s meaning. The doctrine of noscitur a sociis—whereby “a word is known by the company it keeps”—is heavily relied upon by the U.S. Supreme Court to avoid “giving ‘unintended breadth to the Acts of Congress.’” Yet, in the end, many are left wondering why these existential and legal nuances even matter. The Device for the Autonomous Bootstrapping of Unified Sentience (DABUS) saga, currently playing out in both IP offices (IPOs) and court systems around the world, shows why.
Dr. Stephen Thaler, president and CEO of Imagination Engines and creator of the AI system DABUS, began filing patent applications listing DABUS as the inventor in 2018. Thaler also filed a copyright registration application with the U.S. Copyright Office, listing another AI application as that work’s creator. No one denies Thaler is the creator and owner of these AI machines or that this status might grant him certain rights if a patent or copyright registration is approved. Likewise, it is agreed that the innovations in question were made a reality predominantly due to the DABUS AI system.
However, patent law often defines an inventor as “a natural person,” and it’s obvious that AI is not a natural person by any definition of the phrase. Likewise, copyright law requires human authorship. Thus, the great AI-inventorship debate began. What is the definition of a “person?”Is traditional IP law adequate and appropriate for AI systems, or should there be AI-specific reforms? How much human input is required for someone to be deemed an inventor or creator? Can patent applications be submitted without naming an inventor? Should non-legal entities be allowed as inventors? How does this affect other IP? Why does inventorship or creatorship even matter
To unweave this tangled web, one must first understand that IP exists to enable innovation and creation by providing very specific parties with a series of exclusive rights for a limited period of time. Just as employees receive paychecks, benefits, or bonuses as compensation for their work, innovators and creators receive IP rights as compensation. Each of these forms of compensation ultimately feed back into the economy (for example, through individuals purchasing books or patents disseminating information). But just as an employee’s compensation is intended specifically for that employee, IP rights are intended for specific parties in the innovation and creation cycles.
Inventors are one of the parties granted specific rights associated with their patented innovation. Likewise, creators are one of the parties granted specific rights associated with their copyrighted creation. Thus, many patent and copyright laws either explicitly or implicitly require a natural person, or legal entity, to be listed as the inventor or creator. Since an AI system is not a legal entity, it cannot reap the rewards of enforcing, utilizing, or assigning any IP rights. Although Thaler could have certain rights to the patent or copyright as an assignee, no legal entity exists which could undertake the innovator’s rights, including making that assignment. In other words, Thaler cannot obtain rights from DABUS since the AI system could not hold IP rights to begin with. It’s like trying to collect a payment from someone who doesn’t exist.
As of July 13, 2022, seven jurisdictions have weighed-in on the AI-inventorship debate: Australia, the European Union, Germany, New Zealand, South Africa, the United Kingdom, and the United States. The outcome was the same in six of the jurisdictions, although the reasonings varied based on certain nuances surrounding procedural and legal interpretations, as detailed below. It is important to note, however, that all of these decisions are based on interpretations of existing law. Legislative branches could change the law, if they so desire.
This is the DABUS saga:
- In Australia, the patent office initially denied Thaler’s application. The Federal Court then concluded the 1990 Patent Act did not exclude non-humans as inventors. Upon a second review by the court, patentability was denied because a non-human cannot assign its rights, but no decision was made as to whether non-humans could be considered inventors.
- The European Patent Office (EPO) denied patentability on the grounds that the inventor must be a natural person. The EPO Legal Board of Appeal affirmed that decision, stating: 1) the applicant must either be the inventor or “the designation shall contain a statement indicating the origin of the right to the European patent,” and 2) an AI machine cannot assign IP rights.
- The German Patent Office and Federal Patent Court denied patentability on the grounds that a natural person must be named as the inventor. However, the court suggested an application could state an AI machine was involved, in addition to naming human inventors. (This has sparked existential and ethical debates on “who the true inventor is” and how much human input is adequate and appropriate to be identified as an inventor.)
- The IPO of New Zealand denied Thaler’s patent application on the same grounds as the EPO.
- The South African IPO stands alone, granting Thaler’s patent application in July 2021. It is important to note that granting the application was purely procedural, as the South African IPO did not comment on whether an AI machine could legally be an inventor or assign IP rights. Under the country’s depository system, it is still possible the patent could be challenged and ultimately revoked.
- The UK IPO deemed Thaler’s application as “withdrawn for non-compliance” on the same grounds as the EPO. Upon review, the High Court agreed an inventor listed on a patent must be a person, although the justice noted it is literally possible for an AI machine to be the “actual deviser of the invention.” The Court of Appeal, however, was split on the non-compliance issue. While the justices agreed that an inventor must be a person, one justice argued the application should not be deemed withdrawn because the IPO is not required to conduct factual or legal investigations about an application. (This procedural argument, reminiscent of the South African IPO outcome, appears de facto contradictory. It essentially states the IPO can grant a patent if any inventor is listed because the IPO is not required to determine whether a listed inventor is a natural person in accordance with the law.) Thaler is seeking an appeal with the UK Supreme Court. The UK government called for comments on issues surrounding AI and IP, and the UK IPO published updated findings in June 2022. The consultation results indicate that existing patent and copyright laws do not need to be amended given the current status of AI machines, although an exception for text and data mining will be added.
- The US Patent and Trademark Office (USPTO) also denied Thaler’s application for not listing a natural person as an inventor. This decision was upheld by the Virginia Eastern District Court. Judge Brinkema noted in her decision, “[P]olicy arguments do not override the overwhelming evidence that Congress intended to limit the definition of ‘inventor’ to natural persons. As technology evolves, there may come a time when artificial intelligence reaches a level of sophistication such that it might satisfy accepted meanings of inventorship. But that time has not yet arrived, and, if it does, it will be up to Congress to decide how, if at all, it wants to expand the scope of patent law.” Thaler’s appeal on this decision is currently before the U.S. Court of Appeals. Like its UK counterpart, the USPTO is continuously reviewing and seeking comments on issues surrounding AI and IP. ITIF submitted a comment on the matter.
For those keeping score, here’s a chart:
As previously noted, the AI-inventor or creator saga is not limited to patents. Thaler applied for copyright registration on an artwork called “A Recent Entrance to Paradise,” listing “the Creativity Machine” as author of the work and Thaler as the owner. The registration was identified as a “work-for-hire.” The U.S Copyright Office reviewed the application twice before denying it on the grounds that the AI system “lacks the human authorship necessary to support a copyright claim” and Thaler “was not entitled to apply for copyright registration for the work.” On June 2, 2022, Thaler filed suit against the U.S. Copyright Office in the U.S. District Court for the District of Washington, D.C.
Again, why does any of this matter? Because IP exists to enable innovation and creation; if innovators can’t reap the rewards of their labors, innovation is stifled. Likewise, if IP protection is broadly applied, poor quality patents and copyrights might enter the system, also stifling innovation. Why not just leave the inventor or creator line blank? Because someone must hold the rights to IP, and these rights stem from the inventor or creator.
Why not allow non-legal entities as inventors and creators? For the same reason rocks and dogs are not listed; a line must be drawn somewhere, and terms must be adequately and accurately defined to prevent abuse. Furthermore, the issue of whether a non-human creator can own IP was litigated in Naruto v. Slater in 2018, wherein the court held that monkeys are not allowed to claim copyright ownership over selfies they took with a photographer’s camera. These Celebes crested macaques—or black apes—are far more sentient than AI systems, yet neither are afforded the same rights as humans because if they are given legal rights, then they can also violate legal rights. If an ape were to kill a human, it might be killed to protect other humans or moved to a new location. But it would not be assigned a public defender, deposed by legal counselors, and given a trial of its peers. If found guilty it could not pay damages. AI systems are no different.
What about AI-specific reform such as Congress making it legal for robots to own patents? Such reform tends to be limited and lacking foresight, which could cause potential issues as technology progresses into the future. It could also lead to IP protection becoming too broad (enabling poor quality patents) or further narrowed (denying quality innovations of patent protection).
As ITIF’s Center for Data Innovation states, “The role of the patent system is to protect an inventor’s economic rights, not their moral rights. If anything, Thaler’s proposed changes do little to promote innovation—the primary goal of patent law—and instead introduce a legally unpunishable inventor that threatens, not protects, human inventors.” While some countries do recognize moral rights for creators under copyright law, the idea that an AI system could morally object to certain uses of a work is science fiction, at best. Given AI’s lack of sentience, a human element must exist in invention or creation, however minimal it may seem. Under U.S. patent law, an inventor is one who contributed to the conception of an invention, meaning someone who brings a thought or idea into being. A human or team of humans initiate the concept of an innovation then relay that concept to a machine which reduces the innovation to practice (or makes the idea reality). By this definition, an AI system cannot be a legal inventor. The inventor is whatever human first conceives the idea regardless of the role of a computer in helping formulate some output that spurred those thoughts.
The six jurisdictions denying Thaler patents and copyrights based on legal interpretations of rights reserved for natural persons were correct. AI systems are not natural persons, by any stretch of the imagination, and therefore should not be treated as such. Rather, the human conceiving of the innovation or creation (i.e. – the one programming the AI application) should be named as the inventor or creator on patent and copyright applications. Policymakers need to make this clear in their respective jurisdictions so as to create certainty about the role of AI systems in supporting innovation and who benefits from the end results, namely, patents and copyright.