Artificial Intelligence In the Innovation Process: Can AI-Generated Work Be Patented?

With the recent rise of generative artificial intelligence (AI) technology, it’s likely that AI systems will start playing a larger role in the innovation process.

In response to growing concerns around the patentability of inventions created with assistance from AI (including an executive order issued by President Biden in late 2023), the US Patent and Trademark Office (USPTO) recently published “Inventorship Guidance for AI-assisted Inventions,” which discusses how the USPTO will determine inventorship when AI is involved in the innovation process.

Let’s take a closer look at what the guidance currently says — as well as the questions that the guidance has not resolved.

What is generative AI technology?

Generative AI (or gen AI) is artificial intelligence that can produce text, images, videos, and other content based on patterns and structures learned from input data. What it produces is typically in response to user prompts.

How does the technology work? In simple terms, gen AI converts input data into tokens (the smallest unit of information that AI can process), and predicts the next possible token based on the logic of the token sequences it has learned.

Use of the technology is becoming quite widespread — you’ve probably already heard of apps like ChatGPT, Stable Diffusion and Midjourney, and you’ve almost certainly already seen gen AI results for your day-to-day Google searches!

USPTO’s guidance on generative AI: How is inventorship defined?

The USPTO’s guidance stresses that the patent system is meant to recognize and reward human ingenuity. As such, AI-assisted inventions may be patentable, but only natural persons who have made a “significant contribution” to conception may be named as inventors, according to the USPTO’s guidance.

AI cannot be named as an inventor

An important background to the USPTO’s guidance is the 2022 Federal Circuit ruling in Thaler v. Vidal, which held that AI cannot be named as an inventor in a U.S. patent.

In Thaler v. Vidal, Dr. Stephen Thaler had filed a patent application naming his AI system, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), as the sole inventor.

The USPTO originally rejected the patent application on the grounds that inventors must be natural human beings. Thaler then appealed to the Federal Circuit, which upheld the USPTO’s rejection (the Supreme Court declined the case, so the Federal Circuit’s decision is currently the law of the land).

The USPTO guidance on AI-assisted inventions further emphasizes the decision reached in Thaler v. Vidal: inventors and joint inventors must be natural persons.

Inventors must have “significantly contributed” to the invention

According to the USPTO’s recent guidance, a “significant contribution” test is used to determine whether an individual can properly be named as an inventor or joint inventor in a U.S. patent application for an AI-assisted invention. The significant contribution test relies on the so-called Pannu factors — a three-part test first set forth in Pannu v. Iolab Corp.

As laid out in the guidance, the Pannu factors state that each inventor must:

1. Contribute in some significant manner to the conception of the invention
2. Make a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention
3. Do more than merely explain to the real inventors well-known concepts and/or the current state of the art

The guidance further lays out five guiding principles for how the Pannu factors may be applied to AI-assisted inventions specifically:

1. Even if an AI system was used to help create the invention, a natural person can still be an inventor as long as the person’s contribution is deemed significant.
2. A person who merely presents a problem to AI to solve cannot be considered an inventor. However, the way the person constructs a prompt to the AI system in response to a particular problem could be considered a significant contribution.
3. It is not enough for a person to only reduce the invention to practice, meaning that a person who merely recognizes AI output as an invention cannot be named as an inventor. The person must make a significant contribution to the AI system’s output, or conduct a successful experiment using the AI system’s output.
4. A person who designs an essential building block of the invention — for example, building an AI system in response to a specific problem to generate a particular response — can be considered an inventor.
5. A person cannot be named an inventor simply for owning or overseeing an AI system that is used to create the invention.

Without named inventors, an invention is not patent eligible

Under 35 USC § 101 and § 115, any invention that cannot name inventors is ineligible for patent protection.

Accordingly, if there are not any natural persons who made a significant contribution to an invention, the invention is not patent eligible under U.S. law.

Industry responses to the USPTO guidance on AI-assisted inventions

The USPTO guidance has received a generally mixed response from the industry.

On one hand, some agencies have praised the USPTO for providing greater clarity to the role of AI in determining inventorship. For example, one large tech company lauded the guidance for recognizing that AI is a tool for supporting human innovation, and further asserted that “the guidance is broadly applicable to the use of any tool in the innovation process — not just the use of an AI system.”

On the other hand, some agencies have raised concerns that the guidance could stifle innovation in the United States by discouraging the use of AI.

For starters, the USPTO may be holding AI to a higher standard than other tools. The guidance sets up the possibility that, during litigation, an inventor may need to provide evidence for their contribution to an AI-assisted invention.

The American Intellectual Property Law Association (AIPLA) asserts that the “record-keeping” burden should not be elevated for cases related to AI tools. Basically, cases should not experience “wildly different and evolving standards.”

Some agencies have perceived USPTO to be overstepping in its definition of inventorship. The guidance currently frames AI as a potential contributor to an invention. It states: “an AI system — like other tools — may perform acts that, if performed by a human, could constitute inventorship under our laws.”

However, AI arguably cannot conceive of an invention in the same way that a human can; its output is based on mathematical statistics, and it cannot actually understand its own output. Consequently, many agencies have asserted that AI cannot be an inventor by definition. The Institute of Electrical and Electronics Engineers (IEEE) declared that “the USPTO should not place itself in a position to determine if an inventor is human or machine.”

In addition, if only a human is capable of prompting and selecting actual inventions, then the USPTO’s standard for “significant contribution” is insufficient, as it does not account for recognition and appreciation as inventive acts.

According to the American Bar Association (ABA), “Conception is an inherently human activity and is the hallmark of invention. A person must have an original conception of an invention to be an inventor … If an AI algorithm identifies the elements (parts) of something but no one recognizes it … then there is no invention.”

Finally, agencies like the Biotechnology Innovation Organization are worried that the USPTO’s policies could discourage AI adoption — and, given that utilization of AI is growing worldwide, could disadvantage the United States in the global market in the long run.

Navigating the evolving machine learning landscape

The USPTO’s recent guidance has at least established that AI can be used to assist with inventions, so long as the named inventors are natural persons who have made significant contributions.

However, the guidance has not fully clarified AI’s position as a tool in the inventive process, nor has it clearly defined what constitutes an AI tool (as opposed to any other kind of computational tool).

Given that the issue remains contentious, it’s possible that USPTO will continue to update its position and issue further guidance in time to come.

Navigating the patent waters for such a nascent field is exciting, yet challenging. If you’re considering using AI to help with developing your technology, we highly recommend partnering with an experienced attorney early to avoid making any critical mistakes during patent prosecution.

Please feel free to contact us if you have any questions!