Copyright © Françoise Herrmann
If, according to Yuval Noah Harari*, an AI system is the first tool in history to take decisions, execute them, and generate knowledge, in contrast to all previous tools, you might legitimately wonder whether an AI system could be granted patents, just like human inventors. Indeed, couldn’t an AI system generate completely new, useful, and non-obvious things? And in this case, why wouldn’t an AI system be awarded a patent?
For example, Stanford computers trained to learn biology, using a ChatGPT-type program and millions of raw cell biology data, were able to recognize and identify Norn kidney cells, in just six weeks, whereas it took science 134 years to do likewise (Zimmer, 2024). Norn cells secrete the erythropoietin hormone, which stimulates red blood cell production in bone marrow. If an AI system not only sorts and organizes raw data in efficient ways, according to criteria that are provided, but also recognizes and classifies previously unconnected cell data, doesn’t that mean that an AI system will at some point make discoveries that are vital and non-obvious to a biologist skilled in the art, and therefore patentable? That is the question for which the United States Patent and Trademark Office (USPTO) has recently provided Inventorship guidance for AI-assisted inventions (Federal Register).
Inventorship guidance issued following several Requests for Public Commenting (RFC)** on patenting AI-assisted inventions, since August 2019, and more recent meetings with the stakeholder community, as well most recently, President Biden’s Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, on Oct. 30th, 2023. Inventorship guidance published in the United States Federal Register on February 13, 2024, effective for all patent applications (utility, design, and plant) filed before, on, or after that date. Effective with the understanding that USPTO Inventorship guidance is agency policy, without the force of law.
According to Section II of the Inventorship guidance set forth by the USPTO, titled: “Inventors and Joint Inventors Named on U.S. Patents and Patent Applications Must Be Natural Persons” the question of whether an AI system can generate a patentable invention is clearly “no”, even if the use of AI in innovation is largely embraced. The reasons for the clarity of the Inventorship guidance are that, per the specifications of US Patent Law (United States Code,Title 35, Section 100(f)) only an “individual”, understood to be a natural person, can be awarded a patent. This specification of Patent Law also applies to joint inventorship (USC 35, Section 100(g)), which means that even if an invention is AI-assisted, the assisting AI system cannot be considered a joint individual (i.e., a natural person inventor). Conversely, an invention in which AI has played an important role is not un-patentable, as long as one or more natural persons significantly contributed to the invention.
In other words, the current Inventorship guidance does not deny that an AI system could be instrumental in creating an invention. However, the Inventorship guidance clearly specifies that such a system cannot be listed as an inventor because such a system is not a natural person.
Far from definitive, considering the evolving developments of AI, and the ways in which the courts will decide and adjudicate on various cases of inventorship in AI-assisted inventions, the USPTO considers their Inventorship guidance an iterative process, which the public is invited to continue commenting. Specifically, the Inventorship guidance published at the Federal Register specifies:
"The USPTO views the inventorship guidance on AI-assisted inventions as an iterative process and may continue with periodic supplements as AI technology continues to advance and/or as judicial precedent evolves. The USPTO invites the public to submit suggestions on topics related to AI-assisted inventorship so it can address them in future guidance supplements.”
.jpg)
Notes:
*Author of the NYTimes best-selling trilogy Sapiens, Homo Deus, and 21 Lessons for the 21st Century, and Israeli Professor of history at the Hebrew University, in Jerusalem.
**Comments from the RFCs, uploaded to Regulations.gov, may be consulted at: https://www.regulations.gov/. The docket number is PTO-P-2023-0043. Comments from the stakeholder community include, for example, those of Amazon and Google Inc. The comments refer to the USPTO Guidance on inventorship and AI-assisted inventions, published on February 13, 2024 in the Federal Register. USPTO Guidance that was twice opened for public commenting.
References
Biden, J. (Oct. 30, 2023). Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
David, E. (Feb. 14, 2024). Patent office confirms AI can’t hold patents. The Verge.
https://www.theverge.com/2024/2/13/24072241/ai-patent-us-office-guidance
Federal Register – Inventorship guidance for AI-assisted invention.
Harari, Y.N. (2011). Sapiens: A brief history of humankind. New York, NY: Harper Perennial
Harari, Y.N. (2015). Homo Deus: A history of tomorrow. London, UK: Penguin Books.
Harari, Y.N. (2019). 21 lessons for the 21st Century. New York, NY: Random House.
Harari , Y.N. (May 14, 2023). AI and the future of humanity.
USC Title 35, Section 100 (f) and (g).https://www.uspto.gov/web/offices/pac/mpep/mpep-9015-appx-l.html#d0e302338313
USPTO - Department of Commerce (Feb. 13, 2024). Inventorship guidance for AI-assisted inventions. [Docket No. PTO-P-2023-0043]
https://public-inspection.federalregister.gov/2024-02623.pdf
Yuval Noah Harari (website)
Zimmer, C. (March 10, 2024). AI is learning what it means to be alive?
https://www.nytimes.com/2024/03/10/science/ai-learning-biology.html
