The Reality of Using AI in Patent Applications

The rise of AI has left the intellectual property community with an increasing number of questions about how AI can be effectively used in the development and patenting of inventions.

Attribution

Chair Parts by Dennis Scherdt on Unsplash

The 2022 Federal Circuit decision of Thaler v. Vidal established that, in order to obtain a patent, the “inventor” must be a human being. This answered the essential question of whether AI could be listed as an “inventor” on a patent application but raised others about the extent to which a natural person could use and benefit from an AI system during the inventive process. Companies and individuals rely on the use of AI through all stages of creating a patent, from the conception stage to the actual drafting of the application.

There is no question that AI is incredibly beneficial in the processes of patent drafting and prosecution that go beyond the development of the product. The use of AI in writing a patent application allows for the automation of mundane tasks such as checking for compliance, formatting figures, and organizing large amounts of data. It is also beneficial in cost reduction, consistency, and comprehensiveness of searching for prior art. Using AI in the drafting of a patent application is primarily beneficial to the individual inventor, who would be able to reduce the legal costs of patenting by drafting the application via AI. The original policy consideration in the Patent Act, to encourage human ingenuity, still aligns with this democratizing of patent prosecution.

This is not to say that the use of AI in drafting a patent application is not without risks. Accurate disclosures are a prominent issue with AI-drafted applications. Using generative AI can result in claims that are inaccurate. AI hallucination, which occurs when generative AI generates inaccurate or fabricated information, is another issue that often goes unnoticed in patent applications. The solution to these issues is to use AI under supervision and in appropriate, less consequential ways. For example, an attorney can present a written claim to an AI system for proofreading or summary and title generation, without having AI generate the claims itself.

One unique subsection of the use of AI in patent prosecution relates to the AI’s ability to both search for and create prior art. With AI’s expanded search capabilities and broad database knowledge, the search for prior art is far quicker and more comprehensive than a human would be able to achieve. However, this also creates the ability for AI to create prior art, barring future inventors from the use of it. Services are able to use AI to create computer-generated disclosures with the goal of creating prior art. Generative AI is also able to create predictions about future technological developments, which may be considered prior art.

We cannot ignore the reality, though, that many companies rely on AI for the development of their products. For example, over 135 pharmaceutical-based companies benefit from AI in the development and discovery of their drugs, but those companies are unable to obtain patents for the medicines that were developed by an AI system.

The most obvious solution for inventors is to work alongside AI in developing a product. The USPTO’s Inventorship Guidance provides that AI-assisted inventions are not necessarily unpatentable under the holding in Thaler. The statutes on which the decision was based only require the named inventor or co-inventors to be a natural person. The named inventor is determined by who contributed to the conception of the invention. Should multiple inventors have conceived the invention, all involved are named inventors.

A named inventor must have made a “significant contribution” to the invention, which is determined by the Pannu factors. These factors include whether the person conceived of the invention did more than merely explain well-known concepts and contributed more than an insignificant quality or quantity. The Pannu factors have traditionally been applied to human co-inventors. However, they are also applicable in cases of AI-assisted patent applications to determine if the human significantly contributed enough to the conception and development of the invention to be considered a co-inventor.

The use of AI in the creation and patenting of inventions is a widespread phenomenon with little information about what is able to be done and what is not. While more legal regulations are certain to come in the future, inventors are currently able to use AI in many beneficial ways throughout the inventorship process, as long as the AI system is not the named inventor of the product. The USPTO itself has incorporated the use of AI for a more comprehensive and cost-effective search for prior art to compare with applications. But without the proper supervision, AI has the ability to hallucinate incorrect information for an application, flood the patent research with misleading information, and more. Consequently, inventors who use AI to assist with their inventions should be aware of the inherent issues that AI presents to the intellectual property community.