Only people can get patents. There’s a good reason for that, which is that the patent grant—a temporary monopoly granted by the government—is supposed to be given out only to “promote the progress of science and useful arts.” Just like monkeys can’t get a copyright on a photo,  because it doesn’t incentivize the monkey to take more photos, software can’t get patents, because it doesn’t respond to incentives. 

Stephen Thaler hasn’t gotten this memo, because he’s spent years trying to get copyrights and patents for his AI programs. And people do seem intrigued by the idea of AI getting intellectual property rights. Thaler is able to get significant press attention by promoting his misguided legal battles to get patents, and he has plenty of lawyers around the world interested in helping him. 

Thaler created an AI program he calls DABUS, and filed two patent applications claiming DABUS was the sole inventor. These applications were appropriately rejected by the U.S. Patent Office, rejected again by a district court judge when Thaler sued to get the patents, and rejected yet again by a panel of appeals judges. Still not satisfied, in March, Thaler petitioned the U.S. Supreme Court to take his case. He got support from some surprising quarters, including Lawrence Lessig, as noted in a Techdirt post about the Thaler case. 

Fortunately, on April 24, 2023, the Supreme Court declined to take Thaler’s case. That should put an end to his arguments for his AI patent applications once and for all. 

Thaler filed U.S. Application Nos. 16/524,350 (describing a “Neural Flame”) and 16/524,532 (describing a “Fractal Container”) in 2019, and listed “DABUS” as the inventor on both applications. He submitted a sworn inventorship statement on DABUS’ behalf, as well as a document assigning himself all of DABUS’ invention rights. 

“Thaler maintains that he did not contribute to the conception of these inventions and that any person having skill in the art could have taken DABUS’ output and reduced the ideas in the applications to practice,” the Federal Circuit opinion explains. 

But the Patent Act requires inventors to be “individuals,” which means “a human being, a person” in Supreme Court precedent. 

The Idea Of AI Patents Keeps Coming Up

The issue of AI invention won’t go away, because there’s a dedicated lobby of enthusiasts—and  patent lawyers who want to work for them—that wants to keep talking about it. The patent office is currently collecting public comments about the possibility of AI inventorship for the second time, having already done so in 2019

Why would anyone want AI to have inventorship rights in the first place? The amicus brief from a Chicago patent lawyers’ group, which supported Thaler’s case to take DABUS to the Supreme Court, holds a clue. They imagine a future in which: 

ownership can be partitioned in various ways between entities that developed the AI, provided training data to the AI, trained the AI, and used the AI to invent, to the extent that these entities are different. In some cases, such agreements will result in one entity owning 100% of inventions produced by the AI, but other allocations of ownership are possible.

Endless negotiations over slices of idea-ownership might be a win for the lawyers involved in those negotiations, but it’s a loss for everyone else. 

We don’t need property rights systems to govern everything. In fact, the public loses out when we do that. The thousands of software patents created by humans are already a mess, causing real problems for developers and users of actual software. Applications seeking to grant monopoly rights to computer programs created by an AI are a bad idea, which is why we’re giving Thaler’s patent applications our Stupid Patent of the Month award.