Who Owns AI’s IP

AI-created IP ownership considerations

Who Owns AI’s IP

Every once in a while a topic related to intellectual property (“IP”) captures the attention of the mainstream media. Around a decade ago, the popular press got wind of the socalled “monkey selfie” dispute, which centred on whether copyright subsisted in a selfie taken by a macaque and, if so, who owned it (spoiler alert: it’s not the macaque). Now, news articles again circulate on who owns IP purportedly created by someone or something other than a natural person: artificially intelligent machines.

Perhaps primed by Hollywood, the term “artificial intelligence” (“AI”) may conjure images of generally intelligent machines that are able to pass for humans in all manner of circumstances. Reality is less impressive. Society is far from developing an AI that is the stuff of the Terminator or HAL 9000. Rather, task-specific AIs, such as a camera app on a phone that identifies faces or computers that (try) to understand or mimic natural language queries, represent current technology.

Facial recognition and natural language processing are implemented using a type of AI known as a “neural network” that uses an architecture roughly mimicking a human brain to perform machine learning: the ability for a computer to become more accurate in the task(s) it performs based on data without being explicitly programmed to do so. Most people who loosely refer to AI, particularly in comparison to human beings, are referring to machine learning performed with neural networks.

Device for the Autonomous Bootstrapping of Unified Sentience (“DABUS”) is a task-specific AI composed of neural networks tasked with creating new inventions. According to Dr. Stephen Thaler, DABUS’s creator and owner, DABUS invented a “fractal container” and a “neural flame.” Dr. Thaler applied to patent these inventions in several countries listing DABUS as an inventor. While proceedings before the courts or administrative bodies remains pending as of this writing, to date the Federal Court of Australia, the UK’s Court of Appeal, a US District Court, and the European Patent Office have all held that an inventor needs to be a person.

One argument in support of listing DABUS as an inventor is a moral one: someone who creates an invention is simply entitled to be recognized as an inventor. DABUS, however, is something, not someone. As outlined above, like all other AIs, DABUS is a task-specific AI that has a narrow intelligence distinct from that possessed by humans. The day may come when an AI actually does achieve general intelligence and sentience, in which case society will be faced with issues more pressing than whether it should be listed as an inventor on patent applications.

Another argument is an economic one: that if a patent cannot be granted to an AI inventor, then society risks forgoing patent protection for, and consequently the economic incentive to develop, a huge swath of beneficial inventions that may be created by AIs in the future. This concern can be addressed by simply recognizing as an inventor the person who used the AI to create the invention as opposed to the AI itself. It is the person, after all, who invests the time and money to procure, train, and use the AI to create, who should be recognized.

Some countries already apply this type of framework for copyright. UK legislation specifies that in computer-generated works, the author of a work is the individual who made the arrangements that permitted the computer to create the work. A statutory reform for patents analogous to this one for copyright would clarify the legal position of inventors like Dr. Thaler in a manner consistent with the fact that AIs today are just a tool used by creators to whom society owes economic and moral obligations, as opposed to the creators themselves.