Is the "Law of the Horse" finally failing us? Why Canadian IP can’t just "absorb" AI

The emergence of advanced AI and the real possibility of future systems approaching Artificial General Intelligence (AGI) is not merely a new factual scenario for old statutes. It signals a structural challenge to the Canadian intellectual property framework itself.

Canadian legal culture has long been guided by the “Law of the Horse”, Frank Easterbrook’s warning that technology-specific legislation is usually unnecessary. This narrative holds that the Copyright Act and Patent Act can absorb AI just as they absorbed photography, the photocopier, and the internet. That confidence deserves re-examination. Even if AGI does not exist today, the trajectory of current frontier models is forcing us to confront a legal horizon where traditional doctrines of authorship and inventorship begin to lose coherence. At a minimum, Canada must confront how close we are to doctrinal boundaries that were never designed for autonomous creative systems.

The Copyright Fiction: CCH and the Limits of Human Authorship

Canada lacks the explicit statutory “human author” requirement found in other jurisdictions, but the Supreme Court’s jurisprudence, especially CCH Canadian Ltd. v. Law Society of Upper Canada, effectively embedded human intellectual effort at the core of originality – a key component of Canadian copyright. CCH requires that a work originate from an author who exercises skill and judgment, meaning an intellectual effort that cannot be characterized as purely mechanical or trivial.

This standard creates an uncomfortable reality for AI-generated material. When a user provides a prompt and the model generates expression, the substantive creative contribution lies in the model’s statistical processes rather than in a human’s skill and judgment in producing that specific expression. Under CCH, AI-generated outputs therefore struggle to qualify as “original works” because the user’s role lacks the necessary intellectual engagement in the actual execution of the expression. CIPO’s willingness to register some AI-involved works reflects administrative practice rather than a judicial determination, and it does not resolve the underlying doctrinal tension. The result is not a formal assignment to the public domain, but rather a profound uncertainty: many AI-generated materials may be effectively unprotectable under current law. That uncertainty in turn creates a serious investment problem. If the protectability of AI-generated works remains unclear, the economic rationale for investing in advanced generative systems becomes correspondingly weaker.

The Patent Bargain: Strain at Every Seam

The pressures are even more acute in patent doctrine.

The first challenge is inventorship. The Patent Act does not define “inventor,” but Canadian jurisprudence ties inventorship to the individual who contributes to the conception of the inventive idea. Courts have not yet ruled on whether an AI system can be an inventor, but international opinions – such as the global treatment of the DABUS applications – strongly suggest that only natural persons will be recognized. If future AI systems begin generating solutions beyond human conception, identifying a human “inventor” becomes increasingly strained, raising questions about whether the law can continue operating on assumptions that no longer reflect technological reality.

A second pressure point concerns obviousness. The Person of Ordinary Skill in the Art (POSITA) remains central to the Apotex v Sanofi framework. As researchers increasingly rely on AI systems capable of generating technical insights at unprecedented scale and speed, the benchmark of what constitutes ordinary skill may shift, raising the bar for human inventors in ways that distort the patentability threshold. Even incremental evolution in how courts conceptualize “ordinary skill” could have substantial consequences for patent validity across numerous sectors.

Finally, modern AI systems challenge the disclosure requirement. Section 27(3) of the Patent Act mandates that a patent applicant provide sufficient detail to enable a person skilled in the art to reproduce the invention without undue experimentation. Yet frontier models exhibit complex, emergent behaviours that even their developers cannot fully explain or replicate. If an invention’s performance depends on opaque internal dynamics, training data idiosyncrasies, or non-deterministic processes, it becomes debatable whether the disclosure requirement can be met in any meaningful sense. These pressures collectively undermine the integrity of the patent bargain, which rests on a clear exchange between public disclosure and exclusive rights.

The Imperative for Sui Generis Rights

Canada should be cautious about assuming AI is simply another “horse.” As systems become more autonomous, generatively powerful, and capable of contributing to conception – whether full AGI ever materializes – our existing IP frameworks are becomingincreasingly misaligned with technological reality. Attempting to fit AI-generated works into the Copyright Act risks distorting doctrines that were deliberately crafted to protect human creativity. Attempting to force autonomous or semi-autonomous outputs into the Patent Act risks undermining the clarity and coherence of the inventive-concept and disclosure requirements.

A credible path forward is the development of sui generis rights, analogous in spirit to the Integrated Circuit Topography Act, but designed specifically for algorithmic and autonomous outputs. Such a regime could recognize autonomous generation as a distinct category of protectable subject matter, provide shorter terms of protection tailored to the rapid software development cycle, and promote greater transparency around training data and model behaviour. Importantly, it would allow Canada to introduce a principled and technologically informed framework without contorting existing IP doctrines.

Conclusion

Canadian policymakers need not believe that AGI is here today to recognize that our IP statutes were drafted for a world in which only humans conceived and created. As AI systems push deeper into domains once reserved for human authors and inventors, the legal framework must evolve. Canada now faces a moment of choice: adapt our laws to the emerging realities of autonomous generation or continue stretching doctrines beyond their breaking point.

The opinion is the author’s and does not necessarily reflect CIPPIC's policy position.