It goes without saying that generative AI has become ubiquitous in many areas of our lives, permeating everything from art to technical problem-solving to academic research. To put it simply, AI uses an artificial neural network to process large amounts of data and look for patterns, generating output it determines matches what the user has asked. The availability of consumer-use AI software has proliferated in the past couple of years. Users increasingly look to AI to solve problems, and generating a better mousetrap in a few minutes has its attractive aspects.
When the AI generates something that might be patentable, users understandably want to pursue the monopolistic rights available through the patent system. Obtaining a patent can give the owner the sole right to make, use, and sell the invention, with the competitive advantages this may inherently entail. Indeed, generating dozens of innovative and potentially patentable products using AI in the same time that it would take a person to generate one such product seems a no-brainer from a commercial perspective. But can such innovations be patented? Can AI be an inventor, and can the owner of the AI own a patent resulting from the AI output?
Over the past few years, patent offices and courts in many jurisdictions have tackled the issue of AI-generated "inventions". Generally speaking, an inventor is the first owner of patent rights in an invention, and such rights can be transferred to subsequent owners. The issue of inventorship, of who can be an inventor, is thus foundational for determining ownership of patent rights. If there is no inventor under the relevant laws, there is no patent. Some jurisdictions do a better job than others in defining who an inventor is under their legislation and case law, but most recognize an inventor as anyone who contributes to the conception of the invention, or they may make a patentable contribution when reducing the invention to practice. Canadian patent law is aligned with other key jurisdictions in this regard.
But can AI be an inventor? It is generally agreed that AI has been used to generate many useful and innovative products, and with the explosion of consumer-use AI software this will only expand. Efforts have been made to patent such innovations – but (frustratingly or rightly, depending on your viewpoint) those efforts have mostly come to naught.
You may recall over the past few years attempts in numerous jurisdictions by Dr. Stephen Thaler to patent innovations generated by the AI he created, nicknamed DABUS (an acronym for "Device for Autonomous Bootstrapping of Unified Sentience"). It is important to note that Dr. Thaler never put himself forward as inventor or co-inventor, but rather stated that DABUS was the inventor (in Canada, the inventor is listed as "DABUS, the invention was autonomously generated by an artificial intelligence"). Dr. Thaler contended that (a) he created and owned the AI, (b) the AI had generated patentable inventions and should thus be viewed as the inventor, and (c) due to his ownership of the "inventor" Dr. Thaler should be entitled to ownership of the patent rights. There are certainly some interesting points in the various arguments raised around the world in favour of this position (for example, stating that Dr. Thaler would own what DABUS created by the ancient Roman legal principle of accession, akin to owning the calf born of the cow you bought), but almost every patent office and court that has considered the issue has declined to extend inventorship to AI.
Patent applications by Dr. Thaler (as owner – not inventor) have been rejected in many jurisdictions, including the United States, Europe, the United Kingdon, and Australia. At the heart of every rejection is the finding that an inventor must be a natural person – a human. Various administrative and judicial bodies, including our own Canadian Intellectual Property Office, held that a machine cannot be an inventor. While legislation predated the rise of AI systems and thus is understandably of limited use, our own Patent Act refers to an inventor and "his whereabouts", at least strongly suggesting (in gender-archaic terms) that an inventor is a natural person, and patent office and judicial commentary supports this.
This is not, however, the end of the story. While AI-generated inventions have been widely rejected when seeking patent monopolies, AI-assisted inventions are another issue altogether. There appears to be a nascent but growing international consensus that inventions may be patentable if at least one natural person was a true inventor, even if AI was used in some capacity. For example, recently-released guidelines in the United States indicate that you may still have a patentable invention if one natural person "significantly contributed" to the invention. The good news here is that you may be able to seek a patent monopoly even if AI contributed to the invention – but only if the natural person was a true inventor of that invention.
Given the expanding use of AI in various creative endeavours, it is crucial to consider how and to what extent you use the AI in generating potentially-patentable inventions. If you simply feed data and questions into the AI, you very likely will have no claim to inventorship and a patent monopoly will be out of reach. If you instead ensure that you use the AI in only a supporting role – akin, for example, to how modeling or simulation software is used – then you may be able to successfully engage the patent system. But do keep in mind this is not yet completely settled. The mere use of AI when developing an invention could – emphasis on "could" – invalidate a patent claim in some jurisdictions. Every situation must be assessed on its own merits and in light of its own facts, so exercise great caution in mapping out your R&D program to preserve at least the possibility of patenting the results.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
