chevron-down Created with Sketch Beta.

Business Law Today

June 2023

Can an AI Be an Inventor? In the US, the Answer Remains No

Matthew Nigriny and Gary Abelev


  • Generative artificial intelligence is an emerging technology with potentially immense utility, although it raises a slew of important questions and issues related to intellectual property.
  • An individual attempted to secure patent protection on a couple AI-created inventions, but the applications were refused by the U.S. Patent and Trademark Office (USPTO) on the basis that they were not invented by a human.
  • Refusals of applications by the USPTO were upheld by district and appellate courts. Ultimately, the U.S. Supreme Court refused to hear the case, thereby confirming that, for now, patent protection is not available for AI-conceived inventions.
Can an AI Be an Inventor? In the US, the Answer Remains No
Photo by Phillip Glickman on Unsplash

Jump to:

With the recent launch of ChatGPT, artificial intelligence (AI) has been a hot topic in the news. In addition to ChatGPT being a novel and unique tool that may be used in a variety of ways, it creates novel and unique intellectual property issues and concerns.

ChatGPT represents a major improvement in our collective ability to access, process, and convey information. In the past, to answer a question, we would need to perform either simple or extensive research. If the question is relatively basic, a simple search on a search engine might suffice to provide the desired answers. If the question is more complex, additional research might be required, including, e.g., visiting a library and referencing multiple texts. Now, with the introduction of ChatGPT, all that may be required is simply posing the question to ChatGPT. This AI engine is capable of parsing the question presented and providing a thorough answer, regardless of the complexity of the question—though with inconsistent accuracy. 

ChatGPT’s capabilities include improving ease of access to information, forming an appropriate response to the question posed or information requested, and conveying information in a very natural manner. Indeed, if one wanted to research ChatGPT’s capabilities and then write an article about those capabilities, the process could be short-circuited by simply asking ChatGPT to write a short article about itself. ChatGPT can prepare an article that is practically indistinguishable from a human author, without copying prior content. In fact, now you may be wondering if you are being tricked into reading an article written by ChatGPT.

The advancements represented by ChatGPT, however, raise novel issues and present certain concerns. From an intellectual property standpoint, copyright questions abound. For example, ChatGPT is a bit of a black box, and it generally does not include citations or attributions to original sources. Thus, it is unclear if particular content created by ChatGPT could infringe on another author’s copyrights.

ChatGPT may not raise any specific patent-related concerns from a content perspective. However, what if a ChatGPT-like AI was trained to create inventions? That is akin to what Stephen Thaler did, and what subsequently formed the basis for a writ of certiorari presented to the U.S. Supreme Court.

Thaler created DABUS, short for Device for the Autonomous Bootstrapping of Unified Sentience. DABUS is a combination of two AI systems, the first trained with data from a particular scientific area and used to generate novel alterations of that data, and the second developed to measure the novelty and utility of the alterations created by the first.

DABUS created two inventions that formed the basis of two patent applications at the United States Patent and Trademark Office (USPTO). These applications were subsequently rejected on the basis that they were not invented by a human. Thaler appealed to the USPTO’s review board, which upheld the USPTO’s refusal of the applications.

Thaler sought review of the USPTO’s decision in the Eastern District of Virginia. The district court sided with the USPTO, holding that an inventor must be a human. Undeterred, Thaler went to the Court of Appeals for the Federal Circuit (CAFC), which affirmed the district court’s decision—leading to Thaler’s petition for a writ of cert to the Supreme Court.

The crux of the case turned on interpretation of the relevant statutory language. In particular, 35 U.S.C. § 101 states that “[w]hoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor,” while 35 U.S.C. § 100(f) defines inventor as “the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.”

Thaler contended that Congress did not include any restrictions on the words “inventor” or “individual” to pertain to only natural persons. Thus, according to Thaler, DABUS can be an inventor.

The district court held that “individual” ordinarily means a human being. The CAFC held that its prior decisions held that inventors must be natural persons, but acknowledged that the question in those cases pertained to corporations as the competing interest, not an AI program.

In late April, the Supreme Court refused to hear this case. In choosing to not weigh in on whether inventions created by AI may be provided the benefit of patent protection under the laws of the United States, the Court effectively answered that question with a “no,” at least for the time being.

Of course, generative AI is only in its infancy, and the future likely holds additional questions pertaining to the interrelationship between patents and generative AI. Those questions could have wide-ranging implications for both the patent system itself as well as every industry where patents operate.