The Quandary of Patenting A.I. Inventions

24 June 2022Architecture and Engineering, Computational Design, Construction, Digital Transformation, Manufacturing, UncategorizedAI, innovation, robotics



The word “robot” was first used to describe a humanoid in the 1920 fictional Czech-language play “R.U.R.” by Karel Čapek. A general fascination with machines becoming more “human” has been steadily growing ever since.

Today, what’s being referred to as artificial intelligence (A.I.) – including machine learning – is used to discover materials, craft perfumes, develop technology for space travel and self-driving cars, help develop drugs and vaccines, forecast weather, and develop design options for construction. A LinkedIn article relates that computers can talk like people, translate multiple languages, create music, produce paintings, and design products. It follows that inventions involving A.I. could soon touch most aspects of our lives.

Learn all about Computational Design, including A.I., in the new Applied Software Ultimate Guide to Computational Design. Download it for free today.

It seems feasible, then, to copyright or trademark an A.I. system’s creative output. But what about patents? As explained in a commentary, this is becoming a quandary for patent laws around the world.

purple and blue robotic hand touching liquid, creating ripples in pastel colors

Developed in the late 1700s, patent laws assumed inventors would be human. Even when the World Trade Organization finalized its Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS) in 1994, the framers did not foresee that an inventor might be a machine. That same treaty still figures in international patent standards.

Most of today’s A.I. involves machine learning. A.I. development still isn’t at the point where machines can duplicate the diverse, complex thought processes and creativity of people. In construction for example, A.I. and the Internet of Things (IoT) primarily augment or reduce the need for human involvement in a process.

  1. A person sets up the problem.
  2. The person inputs data into the machine.
  3. The machine ranks the data and presents options.
  4. The person interprets and responds to the options.
  5. New data (response) is input to be analyzed.
  6. The machine learns from the response and processes the new data accordingly.
  7. More options are presented for analysis by the person.

The role of the person in the process qualifies them as the inventor.

Copyrights and Trade Secrets

A blog article on points out that, when A.I. is used to help write software, that software can be copyright protected. Since the copyright law requires a human author, having a human review code and make analytical decisions means a human author can be named and the copyright enforced.

Trade secret protection covers information that has economic value as long as it remains valuable and is kept secret. Therefore, the advanced algorithms rooted in A.I., plus inventions that are conceived by A.I., may be protected as trade secrets rather than needing patents.


In its “historical milestones,” the World Intellectual Property Organization (WIPO) explained that about 340,000 patent applications for A.I.-related inventions have been filed since 1956, and from 2012 to now there has been a “patent boom,” with most A.I.-related patent filings in the U.S. and China.

As mentioned in the publication, “WIPO Technology Trends 2019: Artificial Intelligence,” as A.I. develops, hypothetical questions may become real issues, including inventorship. Worldwide, a critical detail of the patent system is who invented the thing being patented; the tech-age question goes further: “Does that “who” have to be human?” Can A.I. systems be listed as inventors on patent filings?


It’s been widely reported that patent applications naming an A.I. system as the inventor have been filed worldwide. In July 2021 the South African Patent Office granted a patent to the A.I. system DABUS as the sole inventor for a food container. Days later, an Australian court did the same, although that was later reversed and is under appeal as of June 2022. As a side effect, this may generate thought around international treaties to protect such patents.

silhouette of person's head with an amber brain against a blue background of light rays and icons

In the U.S. in 2020, as reported by Bloomberg Law and The Verge, the Patent Office ruled under patent law only “natural persons” can be named as inventors:

  • An inventor must be an “individual.”
  • Legal decisions have clarified that individuals are people.
  • A.I. systems are not people.

Many current uses of A.I. don’t create unique inventions that would warrant a patent. Yet, as A.I. technology advances, a machine could devise something “novel” and “non-obvious,” which are specific criteria for a patent.

In most countries for now, those who want patent protection should ensure there is a human involved in interpreting the final results of their A.I. process, so they can be named as the inventor.

Someday A.I. may be sophisticated enough to widely qualify as an inventor. Ultimately it’s up to legislative bodies to decide how patent law needs to change to accommodate A.I. inventions. The commentary in opined that revamping patent law and an international treaty will not be easy, but not doing so will be worse.



New call-to-action

Better Document Management Using the Cloud

26 October 2023Architecture and Engineering, Collaboration, Digital TransformationCarol Dunn

Using the cloud as a repository of information and documentation, it was possible to focus document management on quality rather…

7 Leadership Challenges Involving AE Technology Implementation

12 October 2023AEC, Architecture and Engineering, Digital TransformationCarol Dunn

Leaders who recognize the potential of technology and its impact on their companies are successfully transforming architecture and engineering with…

How to generate EC3 global imperfections on columns

30 August 2023Advance Design, AEC, Architecture and Engineering, Design, Structural designCarol Dunn

According to EC3, in some cases it is necessary to take into account global imperfections in order to consider possible…