Prior Art and AI-Generated Content

Famous musician Nas once said, “There’s nothing new under the sun, it’s not about what you do but how it’s done.” While his quote is partially true, the reality is that patents are granted all the time for new and innovative ideas. But what happens when generative artificial intelligence produces prior art? Does that affect an inventor’s ability to obtain a patent?

The Patent Act of 1790 focuses on three core concepts: novelty, utility, and non-obviousness.  This article focuses on the novelty element and whether content produced by generative artificial intelligence can be considered prior art. If a United States Patent and Trademark Office (USPTO) examiner discovers that prior art exists, the invention is considered to be ‘anticipated’ and will lose eligibility for a patent.

Prior Art Defined

Prior art is an all-encompassing term meant to be a catchall for any previous patent, publication, or use regardless of whether it is domestic or foreign. The eligibility criteria outlined in Section 102 of the Patent Act emphasize the necessity of novelty for an invention seeking patent protection. Section 102(a)(1) specifically states: “Prior art documents and activities which may preclude patentability are set forth in AIA 35 U.S.C. 102(a)(1). Such documents and activities include prior patenting of the claimed invention, descriptions of the claimed invention in a printed publication, public use of the claimed invention, placing the claimed invention on sale, and otherwise making the claimed invention available to the public.”

According to the USPTO: “This ‘catch-all’ provision permits decision makers to focus on whether the disclosure was “available to the public,” rather than on the means by which the claimed invention became available to the public or whether a disclosure constitutes a “printed publication” or falls within another category of prior art as defined in AIA 35 U.S.C. 102(a)(1).”

Below are several landmark cases involving inventions deemed ‘available to the public’ and some of them might surprise you.

• A student thesis in a university library. In re Cronyn, 890 F.2d 1158, 13 USPQ2d 1070 (Fed. Cir. 1989).

• An informational display poster at a scientific meeting. In re Klopfenstein, 380 F.3d 1345, 72 USPQ2d 1117 (Fed. Cir. 2004).

• A document that was electronically posted on the Internet. Voter Verified, Inc. v. Premier Election Solutions, Inc., 698 F.3d 1374, 104 USPQ2d 1553 (Fed. Cir. 2012).

As you can see, the USPTO’s ‘prior art’ barrier is intended to be low and encompass nearly all things “available to the public” which helps to prevent frivolous patents submissions for inventions already in society.

In layman’s terms, for an inventor to fulfill the ‘novelty’ element, they must avoid prior patents, unpublished or published patent applications, any mention in printed publications like articles or technical papers or have public awareness before the filing date of their new patent application. Additionally, the invention must not have been publicly disclosed, used, or offered for sale by the applicant or associates more than 12 months prior to the patent application submission. If any of these conditions are met, the invention is considered “anticipated” and loses eligibility for patent protection. Thus, to qualify for a patent, the machine, manufacture, process, or composition of matter must not have been previously publicly disclosed or known.

Discourse: Electronic Publishing

The major discourse about AI generated content counting as prior art lies within the ‘electronic publication’ portion The Patent Act. Specifically, section 35 U.S.C. 102(a)(1) and pre-AIA 35 U.S.C. 102(a) and (b) which includes any online database or Internet publication such as discussion groups, forums, digital videos, or social media posts.

However, the USPTO claims not all information on electronic publications automatically qualifies as prior art. According to the USPTO “Public accessibility is determined on a case-by-case basis taking into consideration factors such as, where the information is posted, privacy restrictions placed on the posting, the length of time it was posted, and whether the information is indexed for searching.”  

But don’t get your hopes up that easily. An examiner does not need to prove that someone actually looked at the information on social media or in an electronic publication. In fact, “A reference is considered publicly accessible ‘upon a satisfactory showing that such document has been disseminated or otherwise made available to the extent that persons interested and ordinarily skilled in the subject matter or art, exercising reasonable diligence, can locate it.’ ‘If accessibility is proved, there is no requirement to show that particular members of the public actually received the information.’” Jazz Pharm., Inc. v. Amneal Pharm., LLC, 895 F.3d 1347, 1355-1356 (Fed. Cir. 2018) (quoting Wyer, 655 F.2d at 226).

Within the current USPTO guidelines the question remains. Does publicly available information generated by AI, which has been disseminated on social media or otherwise made available via the generative AI platform itself, count as prior art? If we refer to Stare Decisis and Section 102(a)(1) the answer is probably “yes.” But if the answer is affirmative, what does that mean for the future of patent law and prior art?

Generative AI Content Creation

The USPTO has already said that “under current U.S. patent laws, inventorship is limited to a natural person(s).” This stance was upheld by the United States District Court for the Eastern District of Virginia on their decision in Thaler v. Hirshfeld, 558 F. Supp. 3d 238 (E.D. Va. 2021).

The court explained, however, that it was not confronted with “the question of whether inventions made by human beings with the assistance of AI are eligible for patent protection.” Thaler v. Vidal, 43 F.4th at 1213.

Advanced AI algorithms based on generative models like ChatGPT-4 can analyze vast amounts of data and generate innovative solutions that may not have been previously considered by human inventors. Furthermore, the speed at which AI can processes information and create novel combinations of data far surpasses human capabilities. This rapid generation of content raises the question of whether traditional patent systems will recognize all the newly generated content as prior art which has been electronically published.

Current AI Contributions

While it may seem that AI generated content is somewhat alarming to preexisting notions of prior art, it is not all gloom and doom. AI has already made significant positive contributions to the world of science and innovation.

Some examples of this include Google’s ‘DeepMind AI’ which generated 380,000 new materials. Their algorithm was trained a public database of more than 150,000 known materials called The Materials Project. Utilizing this data, DeepMind AI generated designs for over 2.2 million new crystal structures. Almost 380,000 of them were deemed stable enough to proceed to scientific trials for synthesis. This development has substantially increased the total number of known stable materials by nearly tenfold.

Another groundbreaking discovery involved AlphaFold AI which generated all the possible structures for every known protein. The company behind the discoveries made all this information free to download. They even have a claimed goal of making AlphaFold “even more accessible to scientists around the world.” Thus far nearly 500,000 researchers have used the data.

While these innovations are exciting, they raise substantial questions regarding prior art. How can someone patent a new combination of matter if all combinations have already been published? How can pharmaceutical companies patent new drugs if the protein structures vital to their effectiveness are already publicly known? While the USPTO has not created any clear guidelines on generative AI content, they have requested comments for future legislation.

Legislative Challenges

AI-generated content may be considered prior art because it violates much of the criteria for establishing the novelty and non-obviousness of an invention.

1)    Public Accessibility: if AI-generated content is easily accessible to the public online or shared in some other way, it can be treated as “prior art.” Prior art means any existing information or creations that could affect the novelty of an invention. So, if the AI-generated content is out there for the public to see or use, it might be used to question how original or new a later invention really is.

2)    Relevance to the Invention: for something created by AI to be seen as “prior art” for an invention, it needs to be closely connected or similar to that invention. The more relevant and similar the AI content is, the better the case for using it as evidence of prior art against the uniqueness of the invention. With AI content flowing nonstop it may only be a matter of time until every new idea is based on something previously published by AI.

3)    Enablement: AI generated content could count as “prior art,” if it is informative enough for someone skilled in that field to recreate or grasp the invention. If the AI content gives sufficient details and descriptions, it could be seen as prior art because it enables the reproduction of the claimed invention.

4)    Non-obviousness: prior art is not only considered when looking at how new an invention is but also when evaluating how obvious it is. If the AI generated content discusses parts of the invention, and it would be easy for someone skilled in that field to put those parts together, it might affect whether the invention is seen as unique. In other words, it can question whether the invention is obvious or not, impacting its eligibility for a patent.

Conclusion

Deciding if AI-generated material should count as prior art in patent applications will be a delicate balance between the encouragement of innovation and the overwhelming amount of autonomous content produced. Rules regarding AI and prior art are coming soon, but for now policymakers must grapple with the challenges expressed in this article. In the words of famous physicist Stephen Hawking “only time – whatever that may be – will tell.”

Biography

Nathan Kurtin is a 3L who attends University of North Texas Dallas College of Law. As President of the National Emerging Technologies Council and Legal Assistant for the Texas Blockchain Council, Nathan believes that technological innovation is important to the legal field. Upon graduation, Nathan hopes to work in intellectual property litigation where his forward-facing approach and oratorical skills can be put to good use. In his spare time, Nathan enjoys creating art and designing cloths to help benefit various local charities.

For more information about patent applications, see the Klemchuk PLLC Patent Services and Industry Focused Legal Solutions pages.

This article has been provided for informational purposes only and is not intended and should not be construed to constitute legal advice. Please consult your attorneys in connection with any fact-specific situation under federal law and the applicable state or local laws that may impose additional obligations on you and your company. © 2024 Klemchuk PLLC