It is more challenging to determine AI’s place in the legal personhood hierarchy than it might appear. However, the US Patent and Trademark Office announced today that, similar to other forms of intellectual property, the official protections of this type are exclusively granted to individuals. The announcement was made via “guidance,” which is official policy. It is not a hard-and-fast rule. It will shortly fill with the federal register. According to the guideline document (PDF), only “natural humans” are eligible to receive patents due to obvious legal concerns and the basic idea that “patents function to incentivize and reward human ingenuity.”
It is only sometimes evident when analyzing, for example, why firms are considered individuals for certain legal purposes but not for others. They are legal persons. The first modification protects their speech even though they are not citizens and cannot vote.
A legal dispute occurred over whether an AI model could qualify as an individual or if a human was required to award a patent to an “individual.” The guidance states that precedent made it apparent that, unless otherwise specified, an individual refers to a human. However, the issue of whether or how to credit or grant an AI-assisted innovation application still needs to be solved.
Is an AI a “joint inventor” or “coinventor”? For example, if someone created an AI model and that AI model independently created the form and functionality of a patented device? Or does the fact that there was no human inventor in this instance mean that the device can’t even be patented?
AI-assisted inventions are not “categorically unpatentable,” according to USPTO advice. However, AI systems cannot legally be considered inventors because they are not people.
It follows that any given claim must credit at least one human as its originator. Surprisingly, there are significant parallels to the well-known “monkey selfie” case, in which the monkey taking the photo has no right to copyright because monkeys are not legal persons and copyrights are owned by humans. But, they have to demonstrate that they particularly contributed to the innovation, and this is only sometimes easy to do. The way this is defined throughout the document is pretty interesting to read:
Simply identifying an issue, containing an overarching objective, or formulating a research strategy does not qualify as conception. An idea discovered from the AI system’s output may need a proper inventor or joint inventor who is only a natural person posing an issue to the system. However, a noteworthy contribution might be demonstrated by the way the user designs the prompt with a particular issue in mind to produce a certain response from the AI system.
A natural human is only sometimes an innovator if they only identify and value the output of an AI system as an innovation, especially if the output’s characteristics and uses are obvious to someone with ordinary skill. An individual may, nevertheless, qualify as a legitimate inventor if they take the output of an AI system and significantly improve upon it.
Sustaining “intellectual domination” over an AI system does not automatically qualify someone as the creator of any inventions made possible by the AI system.59 Consequently, someone who does not significantly contribute to the invention’s conceptualization but owns or manages an AI system that is employed in its construction can not qualify as an inventor. To put it another way, there’s a type of reasonability criterion at play here that anyone applying for a patent would already be aware of, but there is little precedent for it in the context of AI. It is the rationale behind the guidelines’ existence; people don’t have to worry about whether an AI’s output counts as their innovations because someone “maintains intellectual domination” over it.
The USPTO takes care to clarify that it is not seeking to define or restrict the uses of artificial intelligence (AI) or how it should be applied. It’s merely a new technology applied to an existing statute and precedent. If Congress passes legislation tomorrow declaring AI to be human for IP purposes, the USPTO would “undo” the entire process and create new guidelines for granting patents to AIs. However, until then, artificial intelligence will remain only software, and human labor will continue to be valued and safeguarded.
