October 18, 2023

Inventing with AI

Only inventions conceived by humans are eligible for patents — but humans can use AI as a tool in the process of invention.

AI is increasingly being incorporated into the process of invention, including in areas that involve writing software, developing new materials, and discovering new compounds. This raises complex legal and commercial issues, particularly related to intellectual property and inventorship (for details, see our recent piece on how AI is transforming the drug development process).

It is clear under current US law that inventions conceived solely by AI systems cannot be patented. But it also seems increasingly clear that AI can be used as a tool in the process of invention without rendering an invention ineligible for a patent — if a human conceived the idea for the invention.

In this piece, we explain why and outline the requirements inventions must meet to qualify for patents, even when AI assists in the process.

Ideas solely conceived by AI are not protected under patent law

Patent eligibility in the United States is based on Section 101 of the US Patent Act and related judicial decisions. Lawmakers and the courts have repeatedly indicated that only humans can qualify as inventors.1

The Patent Act expressly states that “[t]he term ‘inventor’ means the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.”2 The Federal Circuit Court of Appeals explicitly stated in 1993 that “only natural persons can be ‘inventors.’”3

Recent decisions applied the principle in cases involving AI. In 2020, the US Patent and Trademark Office (USPTO) denied two petitions by Stephen Thaler that named an AI platform (the Device for the Autonomous Bootstrapping of Unified Sentience, or DABUS) as the sole inventor — precisely because only “natural persons” can be inventors under the US Patent Act.4 The US District Court for the Eastern District of Virginia upheld the USPTO's decision,5 and the US Court of Appeals for the Federal Circuit affirmed the holding.6

In its decision, however, the appeals court stated that the case did not present “the question of whether inventions made by human beings with the assistance of AI are eligible for patent protection.”7

Inventions conceived by humans using AI as a tool likely can be patented

Using AI as a tool in the process of invention should not in itself render an invention ineligible for a patent — if a “natural person” contributed to the conception of the idea. 

There are many ways a natural person could qualify as an inventor or joint inventor, even when AI is used as a tool in the process of invention — including designing the architecture of the AI system, choosing the specific data to input into the AI system, and developing the algorithm the AI system used to process data.

However, the invention still must meet three other main requirements to be eligible for a patent, some of which can be more difficult to meet when AI is part of the process of invention. 

1. The invention must be based on patent-eligible subject matter.

The US Supreme Court developed the Alice-Mayo framework to assess patent eligibility.8 According to the framework, the court first determines whether the claims at issue are directed to a patent-ineligible concept. If they are, the court then evaluates whether the elements of the claim, both individually and “as an ordered combination,” “transform the nature of the claim” into something that qualifies for patent protection.9

Abstract ideas with no clear practical application are not considered eligible. Importantly, software-related inventions, such as those that may be created with the assistance of generative AI, may often been designated as “abstract ideas,” particularly when AI is being used to write code or create software.

To increase consistency and predictability of analysis, the USPTO has established three categories of abstract ideas to consider in prong one of the Alice-Mayo framework:10

  • Mathematical concepts (e.g., a formula or calculation)
  • Mental processes (e.g., a concept performed in the human mind)
  • Methods of organizing human activity (e.g., a fundamental economic principle or commercial/legal interaction)

For inventions that could be considered abstract concepts, it is particularly important to show that the idea is “integrated into a practical application.”

2. The inventor must describe the invention in enough detail that it can be reproduced.

The USPTO may reject patent claims for inventions that cannot be reproduced by a person “with ordinary skill in the art.”11It is incumbent on the patent applicant to describe the invention in enough detail to enable others to reproduce it.

This can pose particular challenges for inventors that use AI, because even computer scientists may not be able to explain why or how an AI system generates particular outputs.

3. The invention must exhibit utility, novelty, and nonobviousness.

An invention has utility if it has a specific and real-world use.12 An AI output that is useful in the present day will satisfy this requirement. 

An invention has novelty if nothing just like it can be found in the “prior art.”13 Inventors using AI in the process of invention may have to take special precautions to ensure that the details about their invention do not become public before a patent is awarded, which could render the invention patent ineligible. This is particularly important for those using AI systems that can be accessed by the public.

An invention is nonobvious if it is significant development over the prior art.14 If the invention is a technical improvement beyond known methodologies, it will meet this requirement.

*      *     *

It will be interesting to see if legislators or courts adjust legal requirements as generative AI becomes more autonomous. But for now, it is clear that inventions created using generative AI can be patented — if a natural person contributes to the conception of the invention, and the invention meets other patent requirements


[1] According to the Dictionary Act, “[i]n determining the meaning of any Act of Congress ... the words ‘person’ ... and ‘individual,’ shall include every … member of the species homo sapiens who is born alive at any stage of development.” 1 U.S.C. § 8(a) (2018).
[2] 35 U.S.C. § 100(f).
[3] Beech Aircraft Corp. v. EDO Corp., 990 F.2d 1237, 1248 (Fed. Cir. 1993) (citing 35 U.S.C. §§ 115-118) (“EDO could never have been declared an ‘inventor,’ as EDO was merely a corporate assignee and only natural persons can be ‘inventors.”’).
[4] In re Application of Application No 16/524,350, Decision on Petition, US Patent and Trademark Office (April 22, 2020), pp 4-5, available at <> [].
[5] Thaler v. Hirshfeld, 558 F.Supp.3d 238 (E.D. Va. 2021).
[6] Thaler v. Vidal, 43 F.4th 1207, 1210 (Fed. Cir. 2022).
[7] Thaler v. Vidal, 43 F.4th at 1213 (emphasis added).
[8] Mayo Collaborative Services v. Prometheus Laboratories, Inc., 566 U.S. 66, 66, 132 S.Ct. 1289, 1303, 182 L.Ed.2d 321.
[9] See Alice Corp. Pty. v. CLS Bank Int'l, 573 U.S. 208, 134 S. Ct. 2347, 2350, 189 L. Ed. 2d 296 (2014)
[10] The draft 2019 Revised Patent Subject Matter Eligibility Guidance, 9-10.
[11] Disclosure refers to patents being awarded in exchange for disclosing the invention through enablement and written description. See Jeanne C. Fromer, Patent Disclosure, 94 Iowa L. Rev. 539, 546 (2009) (explaining that the technical layer in the patent specification describes the invention, that the written description “ensures that the inventor is in possession of the claimed invention,” while enablement requires the patent applicant to “demonstrate in the specification to ‘any person skilled in the [relevant] art [how] ... to make and use the [invention] without undue experimentation”’).
[12] 35 U.S.C. §101.
[13] 35 U.S.C. §102.
[14] 35 U.S.C. §103.