Challenges to Patenting AI-Generated Inventions
With the recent improvements in Artificial Intelligence (AI) technology, the U.S. Patent and Trademark Office (USPTO) has seen an increase in patent applications for AI-assisted inventions. [1]AI helps human inventors gather data, collate it faster, and evaluate outcomes more effectively. At the same time, AI-assistance has created unique issues in patenting inventions that the patent design office has never encountered before.
What is Needed to Patent an Invention
There are three basic types of patents, utility, design, and plant. Design patents are for anything that changes the external appearance of an item, the ornamentation or configuration. Plant patents are for new plants, seeds, or hybrids. Everything else gets a utility patent. A new process, machine, medication, or an improvement to an existing item, goes in the utility basket.
The USPTO has four conditions[2] inventors must meet to obtain a patent for their creation:
- It must be “able to be used.” That is, it must exist and not be theoretical or merely an idea.
- The inventor must have a clear description of how to make and use the invention.
- It must be “novel,” that is, new and never done before.
- It must not be “obvious,” or something anyone could have produced
A fifth requirement is that the inventor or owner must be the one applying for the patent. Joint inventors or corporations who facilitated the invention may apply, but unless the inventor assigned the patent to a third party, nobody else can apply for a patent.
Problem One: Inventorship
On April 22, 2020, the USTPO denied a petition to include an AI system (the Device for the Autonomous Bootstrapping of Unified Sentience or DABUS)[3] as a co-inventor on two patent applications. The decision was upheld on appeal. The Federal Circuit Court held that the Patent Act (35 U.S.C 100(f))[4] states that patents must be given to “individuals” and that other U.S. Supreme Court rulings affirm that “individuals” are “natural persons.” Hence, an AI cannot be granted a patent.[5]
When AI continued to develop, issues of U.S. security came to the forefront. In response to President Joe Biden’s Executive Order on the Safe, Secure and Trustworthy Development and Use of Artificial Intelligence (October 30, 2023), the USPTO has issued and requested guidance documents on how AI-assisted and AI-generated inventions should be treated for purposes of patents. As of this writing, numerous agencies are involved in the attempt to define who and what should be given credit for an invention.
One agency, the Center for Strategic & International Studies (CSIS)[6] has published the results from a workshop that they summarize as:
- Current patent laws are adequate to protect human interests in patent ownership;
- Human contribution and invention should remain paramount in the patent process;
- AI contribution should be acknowledged in the patent
(From CSIS “When AI Helps Generate Inventions, Who Is the Inventor?” February 22, 2024)
Problem Two: Originality
As the world moves into the middle of the 21st century, it becomes increasingly difficult to know if an invention. is genuinely original or “novel” as required by the USPTO. “Novelty” for patent purposes means:
- No one else knew of the invention or used it before the inventor created it;
- No publication in the U.S. or any other country described it before the inventor created it;
- No other individual applied for a patent more than one year before the inventor’s application;
- The invention was not for sale in any public market prior to the inventor’s application.
Novelty has become rare in the global marketplace, simply because so many people have access to the same information at the same time, and many people can have the same idea at once. With tens of millions of individuals sharing information and asking questions on the internet, it will not be surprising if AI-assisted inventions begin to resemble each other.
So far, generative AI such as GPT-4 has not crossed this threshold.[7] Since the generative-AI model is predicated on learning as it is used, inventors and AI users can expect a decrease in “novelty” as more AI-assisted and -generated patent applications appear.
Problem Three: Non-Obviousness
To obtain a patent and the right to be the only individual who owns the new technology, an inventor must show that nobody else could have figured out the invention. That is, it cannot be “obvious.” The patent officer reviewing the patent must determine if any person with “ordinary skill” would consider the invention obvious.[8]
Currently, “obviousness” is determined by what the average person encounters throughout their day. A person with “ordinary skill” can only process a certain amount of data when assembling their invention, so they are limited in what they’re likely to encounter.
Theoretically, an AI of “ordinary skill” has the entire body of whatever discipline is being given it for its invention, hence, nothing would be “non-obvious” according to the current USPTO standards.
The question is not alleviated by AI-assisted inventions. The same issues face inventors using AI to analyze and sort through mountains of online data. Generative AI is beneficial precisely because it identifies connections between disparate data sets. Simply because a human takes those connections and makes them into an invention does not mean that any other similarly situated human would not find the connections equally obvious.
Problem Three(a): Incremental Obviousness
Patents give the patent holder the right to make changes to the invention itself. Product diversification has given us everything from new razor blades to life-saving drugs. Incremental innovation is possible because the patent holder owns the rights to the original invention and can tweak their idea as changes occur to them.
An invention will not be given a patent if it so closely resembles an existing patent that the differences were obvious to a user of ordinary skill. (35 U.S.C. 103)[9] For instance:
- Changing a manual step to automation or computerization
- Changing the composition of a component, for instance metal to plastic
- Changing the number of steps in a process
Sometimes, these incremental changes are not obvious. One strange but true example was a patent for glow-in-the-dark fan blades granted in 2014.
The difficulty AI-generated inventions present is that the volume and speed of their connections means that patent-worthy incremental changes become fewer and farther between. Although glowing fan blades are not necessarily obvious to the average fan user, they may be obvious to enough people that an AI would flag a patent application as “rejected.”
Solutions to the Problem
As it stands, both the United States and the United Kingdom’s patent offices have held that Artificial Intelligence cannot hold patents on its own.[10] Both systems have indicated that AI-assisted patent applications are acceptable and will be judged on their merits.
The USPTO’s guidance[11] on the matter has not yet been tested in the courts, so it remains to be seen how it will function in the real world. The guidance says, among other things, that:
- Inventors and Joint Inventors named on patent applications must be named as “natural persons.” This was the ruling on the Thaler case and prevented the AI “DABUS” from being named as an inventor. All future applications may have only live human beings as applicants.
- AI-assisted inventions are not prohibited from application. AI-assist is still an allowable tool for invention, as long as it is not listed or given as an “inventor” in the application.
- Use of an AI to guide or formulate a research plan may qualify as a “significant contribution” to an invention. The USPTO offers this guidance language: “A natural person who only presents a problem to an AI system may not be a proper inventor or joint inventor of an invention identified from the output of the AI system. However, a significant contribution could be shown by the way the person constructs the prompt in view of a specific problem to elicit a particular solution from the AI system.” (Inventorship Guidance for AI-assisted Inventions, 13 February 2024, page 18)
Inventors and others using AI to help guide their research programs should use an abundance of caution and consult a patent attorney and an intellectual property attorney when filing any patent applications that contain AI-generated or -assisted work. The USPTO has not changed any requirements for requesting supplemental information necessary to evaluate pending applications.
Conclusions
Generative AI is here to stay and is getting more sophisticated each day. Regardless of the Patent and Trademark Office’s opinions, more businesses and corporations are using AI to generate new ideas and streamline old ones, improve efficiency, and spot weak points in existing systems.
For inventors using generative and assistive AI to speed up their invention processes and eliminate duplication, AI is a two-edged sword at best. The same technology that helps improve and refine an invention may keep it from getting the patent the inventors need to protect it from others on the same track.
Keep a close eye on this evolving area of law, and for now, keep artificial intelligence in the background of invention and development.
[1] https://news.bloomberglaw.com/ip-law/ai-use-risks-drop-in-new-patents-as-ideas-are-rendered-obvious
[2] https://www.uspto.gov/patents/basics/essentials
[3] https://www.wipo.int/wipo_magazine/en/2019/06/article_0002.html x
[4] https://www.govinfo.gov/link/uscode/35/100
[5] https://www.federalregister.gov/documents/2024/02/13/2024-02623/inventorship-guidance-for-ai-assisted-inventions
[6] https://www.csis.org/analysis/when-ai-helps-generate-inventions-who-inventor
[7] https://www.investopedia.com/generative-ai-7497939
[8] https://scholarship.law.vanderbilt.edu/cgi/viewcontent.cgi?article=1589&context=jetlaw
[9] https://www.law.cornell.edu/uscode/text/35/103
[10] https://www.wipo.int/wipo_magazine/en/2019/06/article_0002.html
[11] https://public-inspection.federalregister.gov/2024-02623.pdf