In response to an October Executive Order from the Biden Administration, the United Stated Patent and Trademark Office (USPTO) has released its “Inventorship Guidance for AI-assisted Inventions” (“Guidance”). The Guidance sets forth the USPTO’s views on patent inventorship where humans use AI systems to develop their solutions and was released with two example fact patterns that illustrate their views. While there is no categorical prohibition on AI-assisted inventions, the Guidance states that “the inventorship analysis should focus on human contributions, as patents function to incentivize and reward human ingenuity.” Therefore, since AI systems cannot be inventors under the Federal Circuit’s decision in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), a human must qualify as an inventor or joint inventor on each claim in a patent.
The Guidance focuses its analysis on the three factors for joint-inventorship that were set forth in Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998). Under Pannu, a person qualifies as a joint inventor when they: “(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, and (3) do more than merely explain to the real inventors well-known concepts and/or the current state of the art.”
In view of these factors, the Guidance set forth five guidelines for how to determine proper inventorship where a human uses an AI system to develop an invention:
While these positions may not be surprising, there are several important takeaways to consider. For example, the second guideline emphasizes how inventors prompt the AI system in maintaining a human joint inventor. Inventors should take care to avoid simply prompting the AI system to generate a solution to a generic problem. Unfortunately, neither the guidance nor the attached examples provide clear examples of prompts that are constructed “in view of a specific problem to elicit a particular solution” from the AI system. However, iteratively prompting the AI system to refine a solution or significantly constraining the solutions generated by the AI system would likely be helpful in demonstrating significant contribution by a human.
Under the third guideline, a human is not a joint inventor for simply reducing to practice a solution generated by an AI system. But one of the examples provided by the USPTO makes clear that a human can be a joint inventor where a human uses the AI-assisted solution as a starting point and refines the solution based on the constraints of their problem. Additionally, the Guidance clarifies that, in situations where an inventor would not have a reasonable expectation of success of an idea (such as in unpredictable arts), conception does not occur until reduction to practice occurs. For example, a human may still be a joint inventor of a chemical compound provided by an AI system where the human had no reasonable expectation of being able to successfully create the compound.
The fourth and fifth guidelines address the inventorship rights of developers of AI systems that are used to generate AI-assisted inventions. Where a developer creates an AI system that is particularly tailored to solve a problem and someone else uses that system to develop a solution to that problem, the developer would likely be considered a joint inventor on the solution. However, developers of AI systems may not be joint inventors to solutions where those systems are generically applicable or were designed for a different problem. In any case, simply owning or overseeing the system that was used is insufficient for joint inventorship.
As the USPTO has been developing this guidance for inventors, the U.S. Copyright Office has similarly been dealing with the extent to which use of AI impacts authorship. The Copyright Office issued its guidance on this nearly a year ago, similarly emphasizing the need for human authorship and saying the focus should be on whether the AI contributions are the result of mere “mechanical reproduction” or the author’s “own original mental conception to which [the author] gave visible form.” Since then, there have been several related lawsuits in courts as well as decisions by the Review Board of the Copyright Office. In one case, the Review Board was asked to reconsider its refusal to register artwork created by inputting a human’s photograph along with Van Gogh’s Starry Night and asking the AI to apply Starry Night as a “style” to the photograph. The Copyright Office again refused to register the resulting work, saying that there was insufficient creative control by the human. Meanwhile, the Copyright Office requested comments on how it should address such law and policy issues raised by use of AI systems and ultimately received 10,371 comments. Undoubtedly, the Copyright Office will look at all of these developments in updating its guidance on this related issue.
The USPTO Guidance concludes with some reminders to patent applicants of their duties of disclosure and reasonable inquiry. The Guidance states that it is not “changing or modifying” these duties, but makes clear that applicants should provide any information “that raises a prima facie case of unpatentability due to improper inventorship” such as “evidence that demonstrates a named inventor did not significantly contribute to the invention because the person’s purported contribution(s) was made by an AI system.” Additionally, the guidance states that patent practitioners should inquire about “whether and how AI is being used in the invention creation process.” The USPTO Guidance explains that these duties are unlikely to have a major impact, but applicants may consider investigating how AI systems may have been used in developing inventions to ensure compliance with these duties. As with the Copyright Office, the USPTO’s position is likely to evolve as more patent applications with AI issues are examined.