Logo Placeholder

Patent Law Should Permit AI Systems to Be Inventors

by | Sep 7, 2022 | Intellectual Property

This is a continuation of a blog I wrote in February 2020. At that time, no United States court had directly ruled on whether an artificial intelligence system could be an inventor. Now they have, and it’s time to update the story.

Recent Court Decisions: An Inventor Must Be a Person
In September 2021, District Court Judge Leonie M. Brinkema of the Eastern District of Virginia ruled that Congress intended to limit the definition of “inventor” to only natural persons, which means an AI system cannot be an inventor under United States patent law. See Thaler v. Hirshfeld, 558 F. Supp. 3d 238, 249 (E.D. Va. 2021) (citing, among other things, the definition of “inventor” found in the 2011 America Invents Act).

Almost one year later, on August 5, 2022, the Federal Circuit affirmed Judge Brinkema’s decision. Circuit Judge Leonard P. Stark, writing for a three-judge panel, followed essentially the same reasoning as Judge Brinkema and reached the same result. See Thaler v. Vidal, 2022 WL 3130863 (Fed Cir. Aug. 5, 2022) (an inventor must be a natural person).

The reasoning of both judges was straightforward: (1) The Patent Act refers to an inventor as an “individual,” and (2) because courts, as well as dictionaries, have uniformly used the term “individual” to mean a “natural person” or “human being,” then (3) Congress intended inventors to be human beings and therefore an artificial intelligence system cannot be an inventor.

The Patent Act Must Be Updated Now
District Judge Brinkema and Circuit Judge Stark are both rock solid in their analysis of the Patent Act and the meaning of “inventor.” As a result, each of their decisions is virtually unassailable. However, neither decision is satisfying, for the simple reason that they did not solve the problem. The truth is: AI systems actually do invent. And they will continue to invent as AI systems become more and more capable of suggesting new solutions to long-standing problems. If the Patent Act is not at least amended to acknowledge the contribution of AI systems to the inventive process, courts will inevitably be called upon to resolve even more vexing inventor disputes.

For example, I can imagine a creative lawyer finding a way to invalidate a patent on the grounds that the human listed as an inventor is not the actual inventor because an AI system did the real inventive work. One natural result of that argument could be a conclusion that inventorship was improperly credited to the human and therefore the patent is invalid and cannot be enforced. Such an outcome could destroy the incentives to use AI systems to solve difficult problems, which would run contrary to the Constitutional purpose of using patents to “promote the progress of science and useful arts.” U.S. Const. art. I, § 8, cl. 8. But an alternative result of that same litigation could be a determination that the named human was the actual (albeit fictitious) inventor, even though the human took no part in the inventive process. This outcome would create substantial confusion over how a person would qualify as an inventor when an AI system does all the inventive work, not to mention run afoul of the duty of inventors under 37 CFR § 1.63 to affirm, under penalty of perjury, that they are, in fact, a true inventor of the claimed subject matter of a patent application.

I am not arguing that Judge Brinkema’s and Judge Stark’s decisions were wrong. When analyzing laws and the meaning of words used within those laws, judges often view the world through a lens defined by presumptions about what the original drafters were thinking at the time. This is what Judge Brinkema and Judge Stark did, and many people believe this is exactly what judges should do. They believe the job of making new laws is reserved for legislators. They believe judges should only interpret laws that others created, relying on texts that describe the original creators’ attitudes and intentions. It is a judicial philosophy that works as long as laws keep up with the times. The problem, of course, is that things change over time. And if things change enough, it can become difficult for judges to apply old laws to new situations. That is what happened here, and that is why the current law must now be changed.

Patent Law Must Adapt to Reality
AI systems are quickly becoming more powerful and more intelligent. The plaintiff in the Thaler cases (Stephen Thaler) claimed that even though he had developed the underlying AI system, the AI system had been the inventor, not him. What should Mr. Thaler have done in that situation? Should he have claimed he was the inventor? That would have been a false statement, which eventually could have been used against him to invalidate any issued patent.

On the other hand, if Mr. Thaler cannot be an inventor of an AI-created invention, and if an AI system cannot be an inventor, then under the current statutory scheme, AI-created inventions cannot be protected by patents at all. Which means any company that would own an AI-created invention cannot use a patent to prevent others from making, selling, or using that invention, which means the incentives to use AI systems to create novel solutions would be significantly reduced. This seems wildly counterproductive.

There are probably several reasonable solutions to this problem. One fairly easy solution is to amend the Patent Act to let AI systems be inventors, but also to require any resulting patent to be assigned to the AI system’s owner. (This presumes AI systems will never be autonomous and will always be owned, which could be a problematic presumption if AI systems ever become conscious, but let’s put that concern to the side.) Another solution might be to automatically assign an AI-generated invention to its developers, who could further assign the invention to a company, just as human inventors do now. Either solution would be relatively consistent with long-standing property laws. They would incentivize AI innovation. And they would acknowledge the reality of legitimate acts of creativity on the part of AI systems.

Either way, as I argued in my initial article, it makes sense to start developing legal frameworks now that will permit the use of AI systems in the invention process to be incentivized and protected. Otherwise, other countries will take these very reasonable steps (they already have), and our failure to act will slow our ability to compete in the international marketplace.

Clyde Findley is Special Counsel and a registered patent attorney in the Intellectual Property practice at Berenzweig Leonard. He can be reached at [email protected].