No Longer Just a Calculator: Can Machines Be Inventors Too?

Editor’s note: This article is the first in a series of four posts about AI and inventorship, developed as part of Professor Colleen V. Chien’s AI and the Law class at Santa Clara University School of Law. It introduces the upcoming appeal of Thaler v. Hirshfeld, a landmark case asking about AI’s legal place in the international patent regime. The other three posts in the series discuss the technical design of DABUS, how the issue of AI inventorship is perceived by patent agents, and how to determine whether an AI-generated invention is obvious.

This piece was co-authored by Dr. Helen Li (SCU Law ‘13), managing patent counsel at Intel Corporation; Ernie Fok (SCU Law ‘21), patent litigation associate at Weil, Gotshal & Manges; and Akira Kimati, an exchange student at SCU Law.

Dr. Stephen Thaler has challenged courts around the world to answer a question that seems to come straight out of a William Gibson novel: Can artificial intelligence (AI) be an inventor?

Thaler, a physics PhD who also happened to invent the design of the iconic Oral-B CrossAction toothbrush, claims that an AI system he calls “DABUS” has conceived new inventions and art forms, all of its own ingenuity and without his direct input. But when he filed for patents for these inventions on behalf of his AI system with various patent offices around the world, they were denied. For example, the U.S. Patent and Trademark Office (USPTO) denied the applications, holding that only a “natural person” can be an inventor under the law, and that only human beings can be a “natural person.”

The USPTO’s rejection of DABUS’s patents on the grounds that a machine cannot be an inventor is now up for appeal, with the U.S. Court of Appeal for the Federal Circuit set to hear oral arguments on June 6th.

The case invites deep questions about how the law should respond to the rise of machinic intelligence. Is inventorship an exclusively human domain or, if computer programs can invent things too, do we grant them the legal recognition of inventorship? To be sure, Dr. Thaler is not advocating for an AI to own its own patents, as AI does not have legal personality and cannot own property, but for the AI’s owner to own patent’s on any AI-generated inventions.

In an exclusive interview with Santa Clara Law’s AI for Social Impact Blog, the blog of Professor Colleen Chien’s AI and the Law class, Thaler’s attorney Prof. Ryan Abbott urged the public to be open-minded and to think of machines as potential inventors, too.

“We’re moving into a new paradigm where not only do people invent, people build artificial intelligence that can invent. It is important that the law protects AI-generated inventions to promote the use and development of inventive machines that will create socially valuable innovations,” Abbott said.

Following the USPTO’s denial of the DABUS patents, Thaler appealed the decision to the Eastern District Court of Virginia, which upheld the USPTO in Thaler v. Hirshfeld. The appeal of that case is now ongoing under the retitled case, Thaler v. Vidal.

Ryan Abbott and Stephen Thaler. Source: Analytics India Magazine.

In addition to the United States, Thaler has filed his DABUS patent applications in at least 15 other jurisdictions around the world, including the UK, Germany, Australia, China, Korea, and Brazil. His case thus represents a landmark challenge to the patenting regime not only in America, but across the world.

One issue lies at the crux of the legal controversy over his AI system’s patent applications: Can “inventors” be machines, or must they be human? The answer to this question hangs on the statutory definition of the word “inventor.” How other countries have ruled on his case is discussed below, but in the United States, at least, it has been met with a rigid definition limiting the scope of inventorship to human beings only.

In his appellate brief before the U.S. Court of Appeal, Dr. Thaler makes three arguments why the word of “inventor” should be construed to include machines: first, that under its own plain language, the Patent Act refers to terms like “individual” and “person” — and not “natural person” — word choices that indicate that inventorship encompasses a broader category than just humans; second, that denying patents for AI-generated inventions contradicts Congressional intent; and third, that denying them also contradicts the objectives expressed in Article 1, Section 8, Clause 8 of the Constitution, to promote the progress of science.

Specifically, Thaler filed two patent applications with the USPTO that named “DABUS” as the legal inventor of two inventions: a “light beacon that flashes in a new and inventive manner to attract attention” (U.S. patent application Nos. 16/524,350, the “’350 application”) and a “container” based on fractal geometry (patent application Nos. 16/524,532, the “’532 application”). DABUS stands for “Device for the Autonomous Bootstrapping of Useful Information.”


Source: Drawings for Thaler’s patent applications with the EPO.

Figures showing the inventions for a geometric “container” and “light beacon” claimed by the patents.

The problem was not that the two applications failed to satisfy the requirements of the Patent Act for patentable subject matter, novelty, and non-obviousness, but that they named DABUS as the inventor, and DABUS was not a human being (i.e., a natural person).

Why Consider an AI Machine as an Inventor in the First Place?

Under the Patent Act, inventorship relates to who invented the patent’s subject matter. So, whoever contributed to the invention’s “conception,” which is defined as “the complete performance of the mental part of the inventive art,” is deemed an inventor.

When an AI machine alone conceives of an invention, it follows that the AI should be an inventor under U.S. patent law. As modern AI has become more sophisticated with the rise of machine learning and artificial neural networks, AI machines like DABUS can operate in a non-routine, creative manner to produce inventions that go beyond what they were originally programmed for. For a detailed description of how DABUS works, see our second post in the series about the technology.

Thaler recognized that DABUS alone produced the claimed inventions in the ’350 and ’532 patent applications without his direction or instruction. In instances where no natural person has contributed to the invention’s conception, only DABUS should be credited as the inventor on the patent application.

Electro-optical embodiment of DABUS. Source: Imagination Engines.

Then Why Did the USPTO Deny Thaler’s Patent Application?

Section 111(a) of the Patent Act requires the inventor to be listed on a patent application. Thaler thus listed DABUS as the inventor in the ’350 and ’532 patent applications because it was the AI machine that conceived of the inventions.

Though Thaler followed USPTO patent procedure, the USPTO nonetheless rejected his applications, stating that DABUS was not a natural person and that only natural persons can be an “inventor” of a U.S. patent. Thaler disagreed, stating that “inventor” was never limited to only natural persons. He then filed suit in the U.S. District Court for the Eastern District of Virginia.

What Happened in Thaler v. Hirshfeld?

Thaler filed suit pursuant to the Administrative Procedure Act (“APA”), alleging that the USPTO abused its discretion and exceeded its statutory authority by limiting inventorship to natural persons.

The USPTO argued that the district court should defer to the agency’s interpretation of the Patent Act (i.e., inventorship is limited to natural persons). Thaler argued that the USPTO was not entitled to deference because the agency did not provide any evidence that Congress intended to exclude AI-generated inventions from patentability.

In its analysis, the court agreed with the USPTO, but went even further by stating that an “inventor” must be a natural person under the canon of statutory interpretation even if the USPTO’s interpretation is given no deference.

According to the court, the language that Congress used to define the term “inventor” in the Patent Act (35 U.S. Code § 100) — such as “individual” and “himself or herself” — was uniquely trained on human beings. Thaler argued that the term “whoever” in 35 U.S. Code § 101 could be interpreted to mean either a natural person or a machine since the USPTO has interpreted “whoever” to mean either a natural person or not a natural person in the past. For example, 35 U.S.C. § 271(a) uses the term “whoever” to refer to infringing entities. As Thaler stated in his appeal brief “[n]o case states a general principle that language like ‘individual’ or ‘person,’ and subject pronouns must mean a human being.”

However, the district court found a sufficiently dispositive definition of the term “individual” in the Supreme Court case Mohamed v. Palestinian Authority (2012). In Mohamed, the Court concluded that the term “individual” referred to a natural person under its plain and ordinary meaning, and that there is no reason to deviate from such a meaning absent some Congressional intent to give a word a broader or different meaning. Ultimately, on September 3, 2021, the court granted summary judgment in favor of the USPTO, ruling that the inventor of a U.S. patent must be a natural person.

Under appeal, a reversal is unlikely unless Congress amends the definition of “inventor” in 35 U.S.C. § 100. The Eastern District Court of Virginia wrote in its ruling, “As technology evolves, there may come a time when artificial intelligence reaches a level of sophistication that 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.”

A Global Challenge

Thaler’s case has spurred a global debate over when an AI system can be an inventor. The following patent offices and/or courts decided (or at least contemplated) on the inventorship of DABUS patent applications.

European Patent Office — “No”

The European Patent Office (EPO) refused to extend inventorship to Thaler’s DABUS machine. In particular, each application was refused on the ground that it did “not meet the requirement of the European Patent Convention (EPC) that an inventor designated in the application has to be a human being, not a machine.” Thaler appealed the ruling with the Legal Board of Appeal of the EPO, which dismissed the appeal in late 2021, reaffirming that “an inventor designated in a patent application must be a human being.”

United Kingdom — “No”

The two DABUS patent applications filed in the United Kingdom were similarly rejected by the United Kingdom Intellectual Property Office (UKIPO) in late 2019. Thaler appealed the decision to the High Court and subsequently to the UK Court of Appeal. The Court of Appeal issued a unanimous decision in September 2021 that an inventor must be a natural person, therefore excluding DABUS as the inventor of the patents.

South Africa — “Yes”

On July 28 2021, South Africa became the first country in the world to recognize DABUS as an inventor. However, the significance of this decision may be dimmed by the fact that South Africa lacks a substantive patent examination system: as a non-examining country, it automatically grants patents after they are filed.

Australia — “Yes,” then “No”

Two days after South Africa bestowed inventorship on DABUS, the Federal Court of Australia found that an AI can be recognized as an inventor under the Australian Patent Act 1990. As the word “inventor” is not defined in Australia’s Patent Act, the Court considered its ordinary meaning, defining it as an agent noun, which can be a person or a thing (e.g., “computer,” “controller,” etc.). Thus, the Court concluded that nothing in the Patent Act prevents a non-human agent from being named as an inventor.

However, in August 2021, the Commissioner of the IP for Australia appealed the decision, and an appellate panel agreed with the commissioner’s view, stating that only a human can be an inventor. In April 2022, a full court of the Federal Court of Australia unanimously ruled that an artificial intelligence (AI) system cannot be an ‘inventor.’ According to a LinkedIn post by Ryan Abbot an appeal to the High Court is planned.

Germany — A Possible Solution?

In what may be a solution to the conundrum of AI inventorship, the German Federal Patent Court ruled on November 11, 2021 that a natural person must be named as the inventor, but the applicant can state that an AI system was involved in the invention. This holding appears to offer a workaround to the procedural requirement that “inventors” must be natural persons, thereby avoiding the all-or-nothing outcomes of the rulings on the case in the U.S. and U.K.

Canada — Undecided

Canadian courts have not directly expressed their view on extending inventorship to a non-human inventor. However, on November 8, 2021, the Canadian Intellectual Property Office (CIPO) notified Thaler that his patent application did not comply with the Canadian Patent Act “[b]ecause for this application the inventor is a machine and it does not appear possible for a machine to have rights under Canadian law. . . .” CIPO requested a response from Thaler, who asked for an extension of time until August 31, 2022.

As of this writing, Thaler’s DABUS patent applications are still pending in Israel, China, Korea, Japan, New Zealand, Switzerland, India, Brazil and Taiwan.

The Future Is Now

Although the global debate over AI inventorship may have only just gotten started, the proliferation of patents in AI-related technologies is poised to further challenge the current international patent regime. For example, according to an October 2020 report by the USPTO, annual AI patent applications increased by more than 100 percent, from 30,000 to more than 60,000 annually, over the 16 years from 2002 and 2018.

The impressive speed of progress in the field of AI threatens to overwhelm the capacity of the legal system to respond to not only the new innovations created by AI, but the evolution in the very method of inventing itself as AI systems begin to generate entirely new ideas, independently of their human creators. Whether or not the invention of that intellectual property will be attributable to the machine itself remains to be seen.


The opinions expressed are those of the authors and do not necessarily reflect the view of the organizations with which they are affiliated. This article is for general information purpose and is not intended to be — and should not be — taken as legal advice.




The Blog of the Technology, Business, Law, and Policy of AI Class at Santa Clara Law School

Recommended from Medium

Can a Computer Ever Learn to Talk?

Blockchain & AI: How 2 tech buzzwords make sense for the diamond industry

An Oral History of The End of “Reality”

30 Machine Intelligence Startups to Watch in Israel

Legacy of the pandemic

Artificial Intelligence Enables Proficiency

Google Brain -Automating Machine Learning (AutoML)

How AI can Improve the Future Workplace

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


More from Medium

Why is it Exhausting to Let go?

Terra is in a death spiral! The founder announced the plan to issue 1 billion new LUNA tokens!

Why Does Caitlin Johnstone Lie To You?

Joe Walkuski — Industry Veteran of Supply Chain Compliance and Sustainability