The legal status of AI in relation to personhood is not a simple matter (i.e. “none”) — but the U.S. Patent and Trademark Office announced today that, like other types of intellectual property, only a person can get its official protection.
The announcement came through “guidance,” which means official policy but not binding law, and will be published in the federal register soon. The guidance document (PDF) explains that for obvious legal reasons, and also because “patents serve to motivate and reward human creativity,” only “natural humans” can get patents.
It may not be clear when you consider how, for example, corporations are treated as people for some legal matters, but not others. They are not citizens, so they cannot vote, but as legal persons, they have free speech rights under the first amendment.
There was a legal issue about whether, when a patent is assessed for giving to an “individual,” that individual has to be a human, or if an AI model can be an individual. Previous cases showed (the guidance summarizes) that individual means human unless otherwise specified. But it was still an unresolved issue whether or how to acknowledge or award an AI-assisted invention application.
For example, if a person created an AI model, and that AI model designed the form and function of a patentable device on its own, is that AI a “joint inventor” or “coinventor”? Or, maybe, does the absence of a human inventor in this case prevent that device from being patented at all?
The USPTO guidance clarifies that while AI-assisted inventions are not “completely unpatentable,” AI systems themselves are not individuals and so they cannot be inventors, legally speaking. Therefore, it follows that at least one human has to be named as the inventor of any given claim. (There are actually some interesting similarities to the notorious “monkey selfie” case — where the monkey clearly taking the photo can’t get a copyright, because copyrights have to be owned by legal persons, and monkeys, though they are many things, are not that.)
They have to, however, demonstrate that they “significantly contributed” to the invention, and this is not always easy. The document’s explanation of how this is determined is quite fascinating:
Simply identifying a problem or having a vague aim or research plan to follow does not amount to conception. A natural person who only poses a problem to an AI system may not be a valid inventor or joint inventor of an invention found from the output of the AI system. But, a significant contribution could be evidenced by the way the person frames the prompt in light of a specific problem to get a certain solution from the AI system.
…A natural person who only recognizes and values the output of an AI system as an invention, especially when the features and usefulness of the output are obvious to those of ordinary skill, is not necessarily an inventor. But, a person who takes the output of an AI system and makes a significant contribution to the output to produce an invention may be a proper inventor.
Having “intellectual domination” over an AI system does not, by itself, make a person an inventor of any inventions made through the use of the AI system.59 Therefore, a person merely owning or managing an AI system that is used in the making of an invention, without providing a significant contribution to the conception of the invention, does not make that person an inventor.
In other words, there’s a kind of rationality standard at work here that anyone applying for a patent would already know, but which in the context of AI doesn’t have a lot of examples to refer to. It’s for this reason that the guidance exists; no one has to worry now whether, because someone “maintains intellectual domination” over an AI, all its output qualifies as inventions of their own.
The USPTO is cautious to state that it is not in any way trying to define or limit what AI does or is, or how people should use it. It’s simply an application of existing law and precedent to a new technology. If tomorrow Congress passed a law saying AI counts as a human for IP purposes, the USPTO would undo this whole thing and figure out new guidance for giving AIs patents. But until then, AI is still just a piece of software and humans are the ones whose work is meant to be rewarded and protected.