The wait is over – the Canadian Patent Appeal Board (the Board) has answered the question we've been tracking since 2022: can an artificial intelligence (AI) system be named as an inventor on a Canadian patent application? That answer, delivered in Re Thaler, Stephen, is no.
By way of brief background, Dr. Stephen Thaler created an AI system known as "Device for the Autonomous Bootstrapping of Unified Sentience" (DABUS), which allegedly conceived two disparate yet novel products: a food container that enables rapid reheating and a flashing beacon for use during an emergency. Although DABUS itself is the subject of patent protection, Dr. Thaler sought to obtain patent protection over these two products in several countries, and expressly designated DABUS as the sole inventor. The filing documents identified no human contribution to conception.
The Path to Rejection
As we discussed in our February 2022 comment, the Canadian Intellectual Property Office (CIPO) initially flagged the DABUS application as non-compliant, noting that "it does not appear possible for a machine to have rights under Canadian law or to transfer those rights to a human." CIPO suggested Dr. Thaler might remedy this by submitting a statement identifying himself as DABUS's legal representative – a suggestion that ultimately proved futile.
The Board distilled the dispute into three interrelated questions:
1. Can the term "inventor" in the Patent Act and Patent Rules be interpreted to include a non-human entity such as an AI system?
2. Assuming DABUS is not an "inventor," could the application nevertheless proceed without a human inventor?
3. Could any constitutional, policy-based, or international-treaty arguments compel recognition of AI inventorship in Canada?
The Board's analysis rested on three pillars of statutory interpretation: the grammatical and ordinary meaning of "inventor," its context within the Patent Act and Patent Rules, and the purpose of the patent system.
Ordinary Meaning: "Inventor" Means Human
The Board's textual analysis started with dictionaries. While this might seem pedantic, the Board emphasized that the Supreme Court of Canada used the same approach in Harvard College v Canada when interpreting patent terminology. The Board found that "inventor" consistently refers to "one who devises or produces something new," where "one" denotes a person – not any entity, but specifically a natural person.
The Board stated that although a "person" within the meaning of the Patent Act may encompass entities beyond natural persons, such as corporations, the term "inventor" refers exclusively to a natural person.
Dr. Thaler argued that language evolves, pointing to how "calculator" and "computer" once referred to humans but now describe machines. The Board wasn't persuaded. Unlike those terms, there's no evidence that "inventor" has evolved to encompass both humans and machines within the Patent Act context. Ultimately, the Board rejected Dr. Thaler's invitation to read "inventor" expansively to include software as doing so would collapse well-settled distinctions between natural and artificial persons.
Context in the Patent Act Confirms Human-only Inventors
The Board continued its inquiry into the meaning of "inventor" in the Patent Act and Patent Rules, emphasizing that, read in context, Parliament consistently treated inventorship as a human activity.
Although "inventor" is not expressly defined, the language of various provisions in the Patent Act suggests a clear link between a natural person and an inventor:
- Throughout the Patent Act, "inventor" appears alongside distinctly human characteristics.
- Section 31 speaks of an inventor who "refuses to proceed" or whose "whereabouts cannot be ascertained" – concepts incompatible with an AI system.
- The Patent Rules require inventors to have names and postal addresses – requirements that, while DABUS has an acronym and its computer system has a location, weren't contemplated as applying to software or AI.
Importantly, while the Patent Act allows non-natural persons such as corporations to be patentees (patent owners), it maintains a distinction between who can own patents and who can invent. This tracks with the Federal Court's observation in Sarnoff Corporation v Attorney Generalthat Canadian jurisprudence "assumes that an 'inventor' is a natural person as opposed to a juridical person such as a corporation."
The Innovation Bargain Requires Human Participation
The Board also considered the purpose of the Patent Act as it relates to the patent bargain: disclosure in exchange for exclusive rights. This quid pro quo assumes an individual who can both communicate knowledge and be incentivized by the grant. As the Board noted, AI systems "do not need incentives and would not benefit in any way from the reward of a granted patent."
Dr. Thaler invoked the ancient Roman principle of "accession" (the owner of property owns its fruits) to claim ownership of DABUS's inventions. The Board rejected this, finding no precedent for applying accession to intangible intellectual property. The UK Supreme Court reached the same conclusion, distinguishing tangible property (such as a cow producing a calf) from intangible concepts generated by AI.
Policy Arguments Fall Flat
Dr. Thaler warned that denying AI inventorship would chill innovation and leave AI-generated inventions unprotected. The Board remained unmoved, noting that policy considerations don't override statutory interpretation and the evidence didn't support these concerns. The Board noted that pharmaceutical companies have obtained thousands of patents using AI as a tool in development, suggesting no barriers to innovation.
Contrasting the Board's Approach to Canadian Copyright
This decision provides clarity that's currently absent in Canadian copyright law. As we discussed in another past comment, CIPO registered a copyright listing the RAGHAV AI App as co-author of the "Suryast" artwork – a decision now being challenged in Federal Court. The Board does not contend with this apparent inconsistency. As we previously noted, the divergence may stem from human involvement: Suryast had a human co-author while DABUS stood alone.
Practical Takeaways
For innovators and practitioners, the message is clear:
1. Document human contributions: AI can be a powerful tool, but it's important to clearly document human conception and reduction to practice.
2. Name the right inventor: Listing an AI risks application refusal but falsely naming a human inventor when AI autonomously created something could violate section 53(1) of the Patent Act.
3. Structure AI development strategically: Frame AI as a tool that supports, rather than replaces, human invention and human inventors. The thousands of granted patents involving AI tools show this approach works.
4. Watch for legislative change: While the Board won't rewrite the Patent Act, Parliament could. Monitor potential reforms, especially given international pressure and AI's growing capabilities.
5. Consider trade secrets: For truly unsupervised AI-generated inventions, trade secret protection might be the only current option for protection, provided such protection is viable for such an invention.
Looking Ahead – Implications for Innovators & Practitioners
Dr. Thaler can bring an application before the Federal Court to challenge the Board's decision. Given his global campaign for AI inventorship rights (with cases in the UK, US, EU, and Australia), such a challenge might seem likely. But as of now, there's no sign of such a step. In parallel, and as discussed in our past comment , Dr. Thaler and others may lobby for legislative reform.
For now, Canada has drawn its line: inventors must be human. As AI capabilities expand, this bright-line rule provides welcome clarity – even if it means some AI creations remain outside patent protection. Whether Parliament will eventually redraw that line remains to be seen, but the Board made clear that won't happen through creative statutory interpretation.
While many will appreciate clarity on the issue of inventorship, the Board's decision leaves open significant practical questions. How much human input is required to constitute "conception"? How much human input transforms AI output into human invention? Must that input be intellectual, or can it be supervisory? Can someone be an inventor if they merely selected training data or reviewed AI output? These questions will almost certainly arise in future proceedings as AI becomes more prevalent and sophisticated.
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.
