The amount of human involvement needed to secure a patent when artificial intelligence is used to create an invention remains up in the air after a Federal Circuit decision shutting down the possibility of solo AI inventorship.
Patent attorneys expect more litigation on the use of AI in inventions to follow the US Court of Appeals for the Federal Circuit’s ruling last week that artificial intelligence systems can’t be the sole inventors on patents. The three-judge panel noted that the decision was confined to the question of whether computer scientist Stephen Thaler’s creativity machine could be the only inventor listed on a patent application, not whether inventions “made with the assistance of AI” are eligible for patent protection, according to the precedential opinion.
The opinion left unresolved how some provisions of the Patent Act should be interpreted when AI is involved and what constitutes sufficient human contribution for the person to qualify as an inventor, attorneys said. As the US Patent and Trademark Office grants such patents, courts will start having to grapple with new legal challenges surrounding AI inventions across industries.
“The decision leaves some issues open,” said Paul A. Ragusa, an intellectual property partner at Baker Botts LLP. “It’s an area of growing importance.”
New Legal Questions
Thaler’s loss last week marked his latest setback in his quest to convince jurisdictions around the world that his creativity machine called DABUS is the rightful inventor on two patent applications. The Federal Circuit sided with courts in Australia and Europe that found only humans can be inventors under existing statutes. Thaler said he plans to appeal the Federal Circuit’s decision to the US Supreme Court.
It would be up to Congress to change the Patent Act to allow for non-human inventors, but until then, there’s “no ambiguity,” Judge Leonard P. Stark wrote in the opinion.
More challenges to patents created with the help of AI will follow, said Susan Krumplitsch, a partner at DLA Piper, though they likely won’t center on whether the AI should be allowed to be the inventor, as Thaler argued. When inventions rely on machine learning and neural networks, it’s not clear how important the person was in the creation of the invention, she said.
“These issues haven’t been explored,” Krumplitsch said. “I would expect in the coming years, as these patents come up, and we see them in court, and they’re pulled apart, we’ll see more of a focus on who was doing what, and was the human contribution enough to be an inventor contribution.”
If the artificial intelligence system did all or most of the work, the humans involved in the inventions may not be able to take the oath required by the patent office that they are the rightful inventors, said Christopher S. Schultz, a partner at Burns & Levinson LLP in Boston.
‘Bigger Hurdle’ for Other Patents
Patents made with AI may also face greater obstacles to overcome patentability challenges based on obviousness and undue experimentation tests, attorneys said.
Under patent law, inventions can’t be patented if the innovation would’ve been obvious to a skilled expert in the field—called a “person of ordinary skill in the art"—or if it would be too difficult for an expert to recreate the invention.
“If AI is an inventor, then a person of ordinary skill in the art will become a lot smarter,” Schultz said, “which can make the obviousness inquiry a bigger hurdle to patentability.”
Both the obviousness and undue experimentation inquiries are complicated when AI is considered, said William and Mary Law School professor Sarah R. Wasserman Rajec.
“There will be work to be done in patent doctrine about how to treat inventions that are made by a mixture of a natural person and AI,” Rajec said. “What does it mean for something to be obvious if you’re working with AI? What is undue experimentation if you’re working with AI that can run all sorts of experiments and can run things a million times?”
A New Approach?
It’s also unclear how much description is needed to properly show what patent claims were created with AI, Krumplitsch said. The inventions could become so complex that patents may not be the best way to protect them, she said.
Many people still don’t understand what AI is, and how to sufficiently describe the use of machine learning or other aspects of AI to withstand judicial scrutiny hasn’t been tested yet, Krumplitsch said.
Trade secret protection could ultimately make more sense for these inventions, Krumplitsch said. Then, the owners wouldn’t have to face potential patent challenges and their intellectual property wouldn’t be in the public domain, she said.
“It’s a different way of looking at how to protect inventions because it used to be that everything had to be straight up patented, and I don’t think that’s the most prudent way to go forward,” Krumplitsch said. “It’s really changing the bounds of how some of this technology is protectable.”