U.S. Court Rules Artificial Intelligence Cannot be an Inventor (Again)

U.S. Court Rules Artificial Intelligence Cannot be an Inventor (Again)

Sabrina MacklaiSabrina Macklai is an IPilogue Senior Editor and a 2L JD Candidate at the University of Toronto Faculty of Law.


Emily Prieur is an IPilogue Writer and a 3L JD Candidate at Queen's University Faculty of Law.


Artificial intelligence (AI) is everywhere. It influences the ways we socialize, work, and carry out our daily lives. Some even say we are in the AI revolution. But while AI is creating new opportunities and innovations, the law has yet to catch up.

The latest decision from the United States, Thaler v Hirshfeld, comes off the heels of recent judgements in South Africa and Australia asking if AI can be considered the inventor in patent law. This question has sparked widespread debate and commentary within the legal community as legal frameworks struggle to adjust to AI. While South Africa and Australia answered in the affirmative, finding that AI passes the inventor test, the U.S. District Court for the Eastern District of Virginia disagrees.

These decisions can be traced back to a professor at the University of Surrey, Dr. Ryan Abbott. To advocate for AI to be granted inventorship rights, Abbott spearheaded the Artificial Inventor Project. Here, he filed patents for inventions generated by Dr. Stephen Thaler’s “DABUS”, an AI machine that was quite literally invented to invent. South Africa and Australia are just two of seventeen jurisdictions where Abbott and Thaler filed these patents in the hopes of granting inventorship rights for DABUS.

U.S. Rules Against AI as an Inventor

The United States Patent and Trademark Office (USPTO) originally denied Thaler & Abbot’s patent application in 2019, a decision which Thaler and Abbott described as “arbitrary [and] capricious.” The ruling was made on the basis that “no natural person is identified as an inventor.” USPTO supported this finding by citing cases like University of Utah v Max-Planck-Gesellschaft (“Max-Planck”) and Beech Aircraft Corp. v EDO Corp (“Beech Aircraft”) where the Federal Circuit concluded that inventors could only be natural persons.

In the recent 2021 decision, Justice Leonie Brinkema found that the issue of whether AI can be an inventor turns on the plain meaning of the statutory term “individual.” This term is referenced explicitly in the U.S. Patent Act and appears in the America Invents Act’s definition of “inventor” to mean: “the individual, or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.”

The U.S. Supreme Court previously conducted a statutory construction analysis of the term “individual” in Mohamad v Palestinian Authority (“Mohamed”). They ultimately concluded that the ordinary meaning of “individual” referred to a “natural person.” While Congress may intend a different meaning of the term, there must be at least “some indication” of their doing so. Although Mohamad interpreted Congress’ use of the term “individual” in the Torture Victim Protection Act, Justice Brinkema found that the Supreme Court’s analysis remains applicable as the Patent Act does not define the term individual and similarly uses the term as a noun. Therefore, she concluded that since AI is not a natural person, nor normally referred to as an individual, it cannot be an inventor under the Patent Act.

Justice Brinkema also referenced Max-Planck and Beech Aircraft to support her conclusion: “... the unequivocal statements from the Federal Circuit that “inventors must be natural persons” and “only natural persons can be inventors” support the plain meaning of “individual” in the Patent Act as referring only to a natural person and not to an artificial intelligence machine.”

Finally, Justice Brinkema rejected the notion that the court should be the arena to “encourage innovation” and “promote disclosure of information and commercialization of new technologies” through granting patents for AI-generated inventions. Such policy considerations do not override a statute’s plain language. Rather, incentives to create more artificial intelligence machines and other policy matters are problems for Congress to address.

The Future of AI as an Inventor

Abbott is currently appealing this and two other decisions lost at the European Court and in the United Kingdom. In a comment for Bloomberg, he noted that “this decision would prohibit protection for AI-generated inventions and it diverges from the findings of the Federal Court of Australia.” Specifically, the Australian court found that while only a human or other legal person may be an “owner,” it is a fallacy to argue from this that an inventor may only be a human. In defining inventor, Justice Beach found that the court cannot merely resort to “old millennium usages of [the word inventor]” and must recognize the “evolving nature of patentable inventions and their creators.” As such, he found it possible that AI could be an inventor.

As AI grows more ubiquitous, governments will have to intervene and create new legislation that explicitly considers AI and other novel technologies. Otherwise, as noted by Justice Brinkema, courts may be left to make these decisions without proper authority.