Lawyers say overly relying on AI for coding and creative work risks losing protection under Canadian law
If a company uses generative AI to produce a marketing campaign or a piece of software – and a competitor copies it – there may be no legal recourse. At the core of that risk is a question Roch Ripley, an intellectual property (IP) lawyer and managing partner of the Gowling WLG Vancouver office, says Canadian law has yet to resolve: who actually owns the AI-generated content?
“If you prompt it with a very minimal prompt and it outputs something that you end up using, the answer is not you,” Ripley says.
Who owns an AI-generated work?
The ownership question cuts in two directions: who, if anyone, holds rights in the output, and whether that output infringes someone else’s copyright. On the first issue, Canadian law has not drawn a clear line. The US Copyright Office’s first copyright registration for an AI-generated work – a piece called A Single Piece of American Cheese, created using an AI art program called Invoke – was granted after the creator submitted a video of his full process, which involved repeatedly selecting and modifying AI-generated elements through additional prompts, a technique known as inpainting. The Copyright Office found sufficient human authorship in that selection and arrangement, likening the approach to collage. The threshold in Canada remains unclear: “If you put in a prompt saying create a logo for me for my article on Generative AI, full stop… nobody would own it,” Ripley says – with no ownership, there is no standing to sue if a competitor copies that logo.
On the second issue – infringing outputs – litigation is already active internationally. In November 2025, the UK High Court ruled in Getty Images v. Stability AI Limited that Getty’s secondary copyright infringement claim failed: the AI model’s weights, derived from training in the US, did not constitute stored copies of Getty’s images – they were statistical parameters, not reproductions. Getty had limited success only on trademark grounds, for watermarks appearing in early model outputs. Ripley says courts in other countries have reached different outcomes in comparable cases, and the law is often behind in addressing fast-moving technology. “It tends to be [that] the technology and the business [go] first. The law plays catch-up,” Ripley says.
The software question
The implications for AI-generated code came into focus in March 2026, when Anthropic inadvertently disclosed the source code for Claude Code. Third parties republished it, and Anthropic sent Digital Millennium Copyright Act notices – but enforcement was complicated: The head of Claude Code had publicly stated he relied almost entirely on AI to write the company’s software, as Jim Lepore, a partner at Lenczner Slaght LLP in Toronto, and colleagues noted in a May 2026 analysis on AI-generated code and copyright protection. That admission raised the central question: if generative AI developed software substantially or entirely, does copyright protect it?
Lepore, whose practice focuses on IP litigation, notes that copyright disputes over misappropriated code are common. What is new is a foundational challenge: “Does copyright exist at all and is this person who brought this code into existence… properly described as an author within the meaning of the Copyright Act?” he says.
Skill, judgment, and the author question
The Canadian copyright framework derives from CCH Canadian Ltd v. Law Society of Upper Canada, in which the Supreme Court of Canada held that a work must originate from an author exercising skill and judgment, not so trivial as to be purely mechanical. When AI does the actual coding, the question is whether the developer’s prompts clear that bar or amount to mere direction given to the party drafting.
One potential argument against copyright protection, Lepore notes, draws on older Exchequer Court of Canada precedent in Kantel v. Frank E. Grant, Nisbet & Auld Ltd. “The person who actually prepares the work, the person who is hands-on keyboard, historically would be the author of the work, not the person who suggests that work be done or that offers direction,” he says.
The facts of development matter as much as the law – and the records kept may determine the outcome. “[Through] an iterative process, you’re absolutely going to have a stronger argument that you’ve exercised the necessary skill and judgment, and you’re ideally documenting that in a way that when the file comes across my desk as a litigator, I can then point to some of that documentation,” Lepore says. “Documentation is often the least important thing when you’re trying to get a product out the door,” he says – but for companies whose competitive advantage depends on proprietary code, it may matter considerably when something goes wrong.
Lepore does not anticipate that a blanket ruling or statutory amendment will resolve this question definitively in the short term. He expects disputes to be resolved on their facts, and notes the government may prefer judicial input before they decide to change copyright law to address evolving AI technology: “I think that the government might want some of these cases to proceed before the court and get some judicial guidance as to the sufficiency of the existing copyright legislation,” he says.
Building protection before problems arrive
When using AI, Ripley advises clients to distinguish between enterprise tools and public, free versions: when using the latter, “you should presume it’s not [confidential],” he says. Vendors like Microsoft and Adobe offer indemnification for good-faith use, but not for deliberate attempts to produce outputs closely resembling existing copyrighted works. For IP protection, Ripley favours stacking rights rather than depending on any single one: “you could have copyright in the code, potentially copyright in the training data, potentially the code and the data could be trade secrets as well. You could have a patent in the functionality, and you could pair it with like a strong brand…” he says.
Lepore’s advice to software developers is simple: “The first thing is just turn your mind to this. What happens if my code is leaked? What happens if an employee or someone who had access to the code misappropriates it? Am I going to be able to argue that I have copyright protection?” he says. Proactive steps – documentation policies, usage controls, contractual protections – are far less costly than dealing with software source code misappropriation after the fact. “An ounce of prevention is worth a pound of cure down the line,” he says.
