Generative AI and the Future of Patentability
In the ever-advancing world of generative AI, how will the legal system react when future patents are made using AI, especially regarding already patented inventions?
Attribution
Document by RDNE Stock Project
As often happens with the advancement and increasing accessibility of technology, generative AI has begun to intersect with people’s everyday lives. The arrival of this technology has not been universally praised, as some institutions have highlighted the potential issues that the advances in AI will bring relative to market and employment. The International Monetary Fund, for example, has predicted that up to 40% of jobs could be at risk, which in turn could destabilize an already turbulent consumer financial market. As litigation begins in several lawsuits regarding copyright infringement claims with AI, it begs the question of how generative AI will impact other works protected under Intellectual Property law. This blog attempts to predict how courts will rule regarding one specific area of IP: patent protections for inventions.
In order to understand the issues that can come from the use of AI in patents, it is vital to have a basic understanding of how AI currently operates as well as the basics of patent law. AI does not generate information by thinking, empathizing, or using logical reasoning on its own. Rather, it gathers and processes large quantities of data, such that it can produce results similar to the way humans can. A patent grants inventors various rights, including exclusivity in, making, using, selling, offering for sale, or importing the patented invention within the United States for a limited duration. This limited duration is known as a patent term, and they typically last around 20 years. There are three different types of patents: utility, design, and plant. Utility patents deal with new and useful means of process, machine, article of manufacture, or composition of matter, or any new and useful improvements of these. Design patents deal with new and original ornamental designs for an article of manufacture. Plant patents deal with the invention or discovery of distinct and new varieties of plants. All these different patents require that the patent be new and original.
The question, however, then becomes how AI and patentability will intersect. What would be the result if, let’s say, an inventor uses AI to aid and assist him in the creation of an invention that could be patented, but it uses information gathered from other, already patented, inventions? How would courts react? Would the courts consider the use of AI to invent the product so wholly separated from the original creation as to deem it distinct or would the use of another’s patent (or even a small portion of the patented material) constitute an effective infringement of the patent? As of now, there are no notable cases that have tackled these issues. In place of any binding precedent, it might be worthwhile to see how other areas of intellectual property are tackling these issues to better gauge how courts might react in future cases with patents.
One of the most recent (and pending) cases to deal with intellectual property and AI is Andersen v. Stability AI LTD, a putative class action lawsuit in California. That case deals with the defendant, Stability AI, and its generative AI software, Stable Diffusion. Andersen and other artists alleged that Stability “scraped” over 5 billion images from across the internet, including the artists’ copywritten works, to develop Stable Diffusion. As a result, the AI could produce works in the same or similar style of the artist. This case is still pending, however, Andersen prevailed against the defendant’s Motion to Dismiss, meaning that the claim is at least facially plausible against the defendants to move forward with the case. The trial is set for September 2026, and its implications will be colossal. The ruling will likely set the stage for other IP claims against AI developers.
Andersen’s relative success in overcoming the Motion to Dismiss could indicate that a similar claim involving patents might have merit in the courts. It is important to note that, while both are in the intellectual property space, copyrights and patents are inherently different. The Andersen case will likely hinge on whether the AI’s use of her copywritten work is transformative. Patents don’t have a transformative use defense, and most of the common defenses to patent infringement do not deal with using another’s patent to create something new and different like in copyrights. Rather, most defenses tend to hinge on the inventor’s actions during the invention or application for a patent. Given this, if a generative AI model was to use a patented invention (that it scraped from the internet) to create an invention, it is likely that similar litigation would occur. However, due to the defenses to patent infringement being so limited, the owner of the patent may have an easier time asserting their claim. This, however, could change with the ultimate decision of the Andersen case. If the judge or jury sides with the AI developers, it might signal a more lenient approach by courts against AI developers, and the world of patents may very well be included in that. Ultimately, time will be the best indicator of the relationship between AI and patents, but current litigation implies that the patent holder would likely have a successful claim against AI developers.