Takeaways From USPTO's AI-Assisted Invention Guidance
This article was originally published in Law360 on March 8, 2024. Republished with permission.
Pursuant to efforts by the federal government to develop artificial intelligence in a safe, secure and trustworthy manner,[1] the U.S. Patent and Trademark Office issued inventorship guidance for inventions developed with assistance of AI in February.[2]
The guidance clarifies how inventorship is to be determined for the purposes of a patent when AI is involved in the innovation process. It also shows the USPTO’s commitment to adapting examination practices to keep pace with the fast-evolving technological field.
Background
The issuance of the guidance is driven in part from recent attempts to name an AI agent as an inventor of a patent.
Starting in 2019, the Artificial Intelligence Project[3] attempted to obtain a patent listing an AI agent named “Device for Autonomous Bootstrapping of Unified Sentience” as the inventor in a number of jurisdictions, including the U.S.
The USPTO rejected the patent application, declaring that an inventor must be a natural human being.
Stephen Thaler — in his role as representative for the Artificial Intelligence Project — in turn fought the ruling through the U.S. District Court for the Eastern District of Virginia and the U.S. Court of Appeals for the Federal Circuit, both of which upheld the USPTO’s decision of refusal.[4]
At each stage of this ordeal, Thaler had argued that the AI agent should be recognized as an inventor to promote innovation and that the Patent Act does not preclude the listing of an AI agent as an inventor.
Neither the USPTO nor the courts were persuaded by these arguments. The Federal Circuit in particular pointed to a number of sections from the Patent Act emphasizing that the inventor should be a natural person.
Affirming the district court’s decision, the Federal Circuit found that the Patent Act specifically noted that the statute consistently uses the term “individual” when referring to inventors and co-inventors in Title 35 of the U.S. Code, Sections 100(f), 100(g) and 115.[5]
The Federal Circuit, however, also left open the door on “whether inventions made by human beings with the assistance of AI are eligible for patent protection.”
USPTO Guidance
In line with these rulings, the USPTO reaffirmed that AI entities cannot be named as inventors.
Drawing from past jurisprudence on joint inventorship, the guidance specifies that an individual associated with an AI-assisted invention can be deemed an inventor when the individual has made a significant contribution to the claimed invention.[6]
The guidance relies on the Pannu factors — a three-part test articulated in Pannu v. Iolab Corp. in the Federal Circuit in 1998 — in determining what constitutes a significant contribution:[7]
(1) Contribute in some significant manner to the conception or reduction to practice 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.[8]
The guidance applies the Pannu factors to the context of AI-assisted inventions and provides a list of principles to help applicants and examiners determine whether a natural person using an AI system should be listed as an inventor for the purposes of a patent based on the person’s contributions.
Under the guidance, a natural person can be listed as an inventor even if the natural person relied on or used an AI system when the person’s contribution is deemed significant.
Second, a person who merely presents a problem to an AI system and identifies the output from the AI system cannot be considered an inventor; on the other hand, if the person constructed the input prompt to the AI system in a particular way to elicit a particular solution, the person’s contribution could rise to the level of significance for inventorship.
Third, a person who merely reduces an invention to practice alone is not a contribution that rises to the level of inventorship. For example, someone who appreciates the output of an AI system especially when the “properties and utilities are apparent to those of ordinary skill,” is not necessarily an inventor.
Fourth, a person who designs an essential building block, such as building or training the AI system in view of a specific problem to elicit a particular solution, can be considered to have provided a significant contribution.
Fifth, an individual who maintains intellectual domination over the AI system does not — on its own — make the individual an inventor of any of the inventions created in conjunction with the AI system.
The guidance notes that if no natural person has made any significant contribution to the claimed invention, then no inventors can be named, and the application should be rejected on Title 35 of the U.S. Code, Sections 101 and 115, grounds.[9]
Examples Provided by USPTO
In furtherance of the guidance, the USPTO also provided two examples illustrating how inventorship should be determined for claims related to AI-assisted inventions.[10]
The first example provides a narration of how a transaxle for a toy remote control car was created.[11] Here, two natural persons rely on an AI system to create a preliminary design for a transaxle and appreciate that the output design could be used in the remote control car.
This example presents five scenarios, each with varying levels of human involvement in the conception of the transaxle design:
1. Natural persons take output from an AI system without any alteration.
2. Natural persons make minimal alterations to the output of the AI system, while reducing the transaxle to practice.
3. Natural persons perform experiments on the AI output to create a modified design.
4. Natural persons use an AI system to make minor alterations to a new design that they came up with.
5. The owner of an AI system attempts to patent a transaxle design.
In line with the USPTO’s own guidance, the example explains that the individuals under the first and second scenarios cannot be considered proper inventors under the first through third guiding principles laid out above, because they did not make any significant inventive contribution other than appreciating that the design would work and reducing the design to practice.
On the other hand, the example lays out that the individuals under the third and fourth scenarios can be considered proper inventors for the purposes of a patent.
In the third scenario, the example highlights that these natural persons made significant contributions by conducting experiments to see how to modify the original design and that these modifications were integral to the claimed invention.
In the fourth scenario, the example notes that the use of the AI system to modify a new design does not negate the individuals’ contributions as inventors.
As for the fifth scenario, the example notes that the owner of the AI system, which the two individuals used, cannot be considered an inventor for the patent only on the ground of ownership.
This example puts forth a concrete, real-world example of what to consider when deciding whether to name an individual as an inventor on a patent.
Conclusion
The new USPTO guidance regarding AI and inventorship is set to shake things up for patent practitioners.
The patent practitioners will need to get up to speed on the specifics of the guidance and how it applies to their clients’ inventions.
This includes advising clients on how to document the inventive process in a way that meets the USPTO’s new requirements. For example, merely presenting a problem to an AI and acknowledging its initial output will not qualify one for inventorship.
Instead, inventorship will require conducting experiments, modifying the AI’s output or providing essential building blocks for the claimed invention, among other actions, to make a significant contribution.
The USPTO guidance also emphasizes the importance of the duty of disclosure and the duty of reasonable inquiry. In this regard, patent practitioners will need to adjust their practices to accommodate the new guidance.
For example, under the duty of reasonable inquiry, patent practitioners should take keen notes on the facts and circumstances surrounding the process leading up to the conception of an invention, especially when an AI system is involved with a portion of this process.
Practitioners should be aware that:
- The mere use of an AI system does not negate an individual’s inventive contribution.
- The presentation of the problem to the AI system and recognition of the use of the output from the AI system does not on its own constitute significant contribution.
- Reduction of the invention to practice is not sufficient for inventorship.
- The development of an essential block for the claim invention could constitute significant contribution.
- Ownership or intellectual dominion over an AI system on its own does not confer the person the title of inventor.
The new guidance should be seen as a positive step to boosting the burgeoning, nascent field of AI, with a global market for AI expected to reach a staggering $1.8 trillion by 2030.[12]
The guidance regarding the use of AI systems to assist in providing technical solutions to technical problems aligns with the long-standing principle that inventors must be natural people.
The principles laid out in the guidance are not a deviation from legal precedent, but rather a clarification applied to the rapidly growing field of AI-assisted innovation. The USPTO’s guidance brings clarity to the issue of inventorship in the age of AI-assisted inventions.
This clarity should help ensure patents are granted to the rightful inventors, those who make significant contributions to the inventive process.
Navigating this terrain, however, comes with its own hurdles. Figuring out what exactly counts as a significant contribution when AI is part of the invention process is not always clear-cut.
For example, the new guidance could spark controversies regarding who rightfully deserves credit for an invention, particularly in cases where AI systems play a pivotal role.
While there are some uncertainties, the immediate risk of prosecution for failing to meet the new inventorship standards appears low. The extent to which examiners will scrutinize AI involvement in inventions remains to be seen.
All in all, while AI will significantly accelerate the pace of innovation, benefiting everyone, it is human ingenuity and creativity that will continue to drive invention and patenting activity for the foreseeable future.
[3] https://artificialinventor.com/.
[4] Thaler v. Hirshfield 
, 558 F. Supp.3d 238 (E.D. Va 2021); Thaler v. Vidal , 43 F.4th 1207, 1213 (Fed. Cir. 2022), cert denied, 143 S. Ct. 1783 (2023).
[5] Thaler v. Vidal at 1211.
[6] 89 Fed. Reg. at 10047.
[7] Pannu v. Iolab Corp. , 155 F.3d 1344, 1351 (Fed. Cir. 1998).
[8] 89 Fed. Reg. at 10047 (quoting HIP Inc. v. Hormel Foods Corp. , 66 F.4th 1346, 1353 (Fed. Cir. 2023)).
[9] https://www.law.cornell.edu/uscode/text/35/115 (on inventor’s oath or declaration).
[10] https://www.uspto.gov/initiatives/artificial-intelligence/artificial-intelligence-resources.
[11] https://www.uspto.gov/sites/default/files/documents/ai-inventorship-guidance-mechanical.pdf.
[12] https://www.statista.com/outlook/tmo/artificial-intelligence/worldwide.
