Questions about AI inventorship remain after the EPO publishes its justification for refusing two patent applications for machine inventors
Reading time: five minutes
The EPO’s justification for refusing a patent application that had an artificial intelligence tool listed as the inventor does not provide enough clarity on the rules surrounding AI-generated inventions, according to in-house and private practice IP lawyers.
Last month, the EPO published its grounds for denying patent applications citing an AI system named DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the inventor. One reason for the refusal was that an inventor must be a “natural person”, as set out by the European Patent Convention. Another was that AI cannot own the rights to a patent because it has “no legal personality”.
The published justification leaves room for interpretation over how much involvement a human can have in creating an AI-generated invention.
Gareth Jones, vice president of IP at Benevolent AI in London, says he believes the DABUS applications show that machines are now at a point where they are inventing on their own. This development is likely to lead IP into new legal territories, he says.
“I have heard that patents have been granted where machines were actually the inventors but humans were listed. We are at a point where machines are inventing. But the question is: on what scale is that happening?
“This isn't simply technology supporting an invention, where the human provides the inventive contribution. We're not talking about machines thinking for themselves either. This is technology making contributions we'd consider inventive if a human had done it,” says Jones.
According to a Managing IP survey taken of in-house counsel last year, just over a fifth of respondents (22%) said that AI is already used for invention discovery. Of those respondents, three quarters reported either not disclosing or knowing the role of AI when they filed the patent application.
Not there yet
Claire Bennett, general counsel at Five AI in the UK, says that she does not believe AI has the general intelligence needed to invent just yet, but she is curious to see how the law will develop once machines become sophisticated enough to create inventions without human intervention.
She adds: “The more interesting question is: which entity or entities involved in the process of generating the invention should be named, and to what extent should those involved in developing the AI be credited as inventors?”
For Matt Hervey, head of AI at law firm Gowling WLG in London, the published decision leaves several questions unanswered. He says that he does not see an express statement in the judgment that there was no patentable invention in the application. This omittance leaves the lingering question of what applicants should do when there is no human inventor.
“Where AI has been involved, what happens when a human inventor has been wrongly named? Let’s say in the DABUS example, they named a person who co-wrote one of the applications, Ryan Abbott. What would have happened?
“Under European procedure, only the true inventor can object to a false claim, but what if that inventor was a machine?” Hervey continues.
“The decision does not address how to assess the human contribution to the invention. The machine didn’t turn itself on or gather its own data.”
Abbott speaks out
Applicant Ryan Abbott, professor of law and health sciences at the University of Surrey, says the EPO’s decision to deny AI inventorship threatens the integrity of the patent system and the moral rights of human inventors because it invites people to take credit for work they have not done.
“What is important is that the EPO decision would effectively create a new condition for patentability that was never intended by prohibiting protection for inventions made without a traditional human inventor,” he says.
“The decision did not articulate rules for who should qualify as an inventor for an AI-generated invention and under what circumstances. This would seem to prohibit patents on AI-generated inventions unless someone was less than forthcoming about their involvement.
“We do not think this is the right approach because it threatens the integrity of the patent system and encourages dishonesty among applicants,” he adds.
However, an EPO spokesperson says that these arguments go beyond the scope of the decision, which is based on applicable legislation and which may still be subject to an appeal: "The office closely follows the discussion of AI patentability in society ... These discussions often look at questions such as rights to be possessed by machines and the automatic transfer of rights to an owner of a machine or of data."
Nikolas Kairinos, CEO and founder of AI company Fountech in Cyprus, agrees with the EPO’s justification, and says he believes it is nonsensical to give AI inventorship status without considering the wider legal issues that it would create.
“Will AI have the same rights as a human inventor? Can it claim royalties? Could it sue a person who copies the invention? There are simply too many unanswered questions in need of addressing before we can make such a monumental judgement.”
He adds: “AI is undoubtedly a useful tool, which, when working alongside humans, will inevitably contribute to the creation of some remarkable inventions. However, we can’t simply jump at the chance to break records and award AI with patents without thinking about the knock-on ethical and legal impact.”
As one of the reasons for its decision (point 27), the EPO wrote that AI cannot have rights because it does not have a legal personality comparable to natural or legal persons.
“Legal personality is assigned to a natural person as a consequence of their being human, and to a legal person based on a legal fiction. Where non-natural persons are concerned, legal personality is only given on the basis of legal fictions. These legal fictions are either directly created by legislation, or developed through consistent jurisprudence. In the case of AI inventors, there is no legislation or jurisprudence establishing such a legal fiction.”
As a result, the EPO added, AI systems or machines cannot have rights that come from being an inventor, such as the right to be mentioned as the inventor or be designated as an inventor in a patent application.
But the DABUS team says it is not arguing that AI should own the rights to the patents, but rather that the tool should be given the status of inventor.
Abbott at the University of Surrey says: “It is unfortunate that many of the objections to our test cases focus on this non-issue.
“Most inventors do not own their patents, and we argue that an AI's owner should own any patents on its inventions, the same way he or she would own a trade secret generated by AI or any property created by their machine.”
Jones at Benevolent AI agrees that there is confusion between ownership of patent rights and inventorship: “People confuse ownership with inventorship. But ultimately ownership is far more important than inventorship.”
The DABUS team says it plans on appealing the decision to the EPO’s Boards of Appeal. In the meantime, what is certain is that as technology advances and AI becomes more sophisticated, the grey areas of the law around AI inventorship will need more clarity.