US patents boss cannot stress enough that inventors must be human, not AI

You can use neural networks to help, sure, but not do all the work

The US Patent and Trademark Office this week repeated loud and clear it will only accept patent applications that list actual real humans as the inventor and not AI.

That said, officials signaled they are willing to consider inventions developed with the assistance of artificial intelligence, though the human inventors listed on the submitted paperwork must have made a "significant contribution" to the blueprints.

Really, not much has changed: Inventors and joint inventors must be natural persons, and those persons must pass the standard Pannu test, which among other things requires them to have made substantial contributions to the designs.

That's pretty much been the position of the USPTO and the courts ever since certain folks started trying to apply for patents with AI or software listed as the inventor: US law is clear that inventors must be human, and they must have pulled their weight.

What's new is that Kathi Vidal, Under Secretary of Commerce for Intellectual Property and Director of the USPTO, has issued detailed guidance [PDF] reiterating the above while clarifying that inventors have room to use AI to aid the development of a patent application.

Machines can even be credited and thanked in the submission, but not listed as an inventor. In the same way inventors are allowed to use calculators, CAD packages, and suchlike, they can use artificial intelligence; ultimately the final submission must be significantly human-made, and name those human inventors, if it's to be accepted.

The right balance must be struck between awarding patent protection to promote human ingenuity and investment for AI-assisted inventions

"The patent system was developed to incentivize and protect human ingenuity and the investments needed to translate that ingenuity into marketable products and solutions. The patent system also incentivizes the sharing of ideas and solutions so that others may build on them," she said in a separate memo.

"The right balance must be struck between awarding patent protection to promote human ingenuity and investment for AI-assisted inventions while not unnecessarily locking up innovation for future developments."

The USPTO has previously rejected attempts to patent inventions assigned to software. That positioned was challenged in multiple legal appeals led by Stephen Thaler, an entrepreneur advocating for the formal recognition of computers as inventors. Thaler even tried to take his fight all the way to the US Supreme Court, to no avail. 

Interestingly enough, a detailed prompt for an AI system, asking it to solve a particular problem, may be considered a significant contribution by a human inventor, though it seems simply posing basic questions to a neural network and funneling the output into a patent submission is not enough. You have to do more than tell an AI to do your job for you as an inventor. A decent amount of effort should go into shaping the model's output and/or interpreting or developing the results.

"If an individual made a significant contribution through the construction of a prompt, that could be sufficient. However, maintaining 'intellectual domination' over an AI system, does not, without more, make a person an inventor of any inventions created through that AI system," Vidal clarified.

Owning or operating an AI system isn't enough, either.

The USPTO will not require inventors to specifically disclose that they used AI in their work, mainly because, it seems, that would be a pain to define and assess.

"As AI becomes ubiquitous, including as people build on each other’s AI-assisted inventions, it will become increasingly difficult to identify the ways in which AI plays a role in the inventive process. Therefore, the USPTO is not, at this time, implementing any new requirement to disclose the use of AI, beyond that which is required in rare circumstances by USPTO rules," Vidal said.

That all said, the USPTO boss warned this doesn't mean humans will be automatically entitled to patents designed with the help of AI. Officials will examine, as usual, whether a human really contributed significantly to a particular invention, and whether other requirements are met, before considering granting an application.

"From an examiner's perspective, it will not matter if AI, or other advanced computer system, performed actions that would rise to the level of inventorship. What matters, under the guidance, is whether at least one human's actions can be shown to rise to the level of inventorship and is listed as an inventor on the application," Vidal said.

Other criteria have to be met to successfully get a patent approved, again as usual, including: subject matter, utility, and novelty, according to the latest guidance from the USPTO. ®

More about


Send us news

Other stories you might like