There’s no stand-alone “deepfake law” in Canada — but how you use one can already break several existing rules.
Short answer
Making a deepfake isn’t automatically illegal on its own; Canada has no dedicated statute that names the act. But the federal government treats synthetic media as a named risk category, not a settled legal question, and several existing laws — deceptive-marketing rules, and the ordinary tort of passing off — already reach specific ways a deepfake gets used.
The Canadian Artificial Intelligence Safety Institute, part of the federal government’s AI safety work, names the risks it exists to study, and deepfakes are explicitly on that list: “risks posed by synthetic content, including impersonation and fraud.” That’s a policy body identifying a hazard, not a court naming an offence.
The Canadian Centre for Cyber Security groups the same problem under a broader heading. Its guidance on generative AI lists “misinformation and disinformation” among the technology’s eight named risks, and warns generally that AI output “can be incorrect” and “you should always be aware of and validate your sources to verify whether the content being presented is accurate.” Neither document creates an offence; both confirm the government is watching the problem closely enough to name it twice.
The clearest hook is consumer-protection law. The Competition Bureau’s own discussion paper on AI names the exact mechanism: AI “could be leveraged in deceptive marketing practices,” including through “‘deepfake’ (digitally altered) voices, images, or videos,” and “AI eases the ability to scale such conduct.” A deepfake used to fake a celebrity endorsement or a fraudulent prize notice falls straight into Canada’s existing false-advertising rules.
Where a deepfake creates confusion about which business it’s really connected to — a fake video implying a competitor’s endorsement, say — the ordinary common-law tort of passing off can apply the same way it would to any other brand impersonation. The elements are unrelated to AI: you generally need to show your business has genuine goodwill in its name or branding, that the confusing use is likely to deceive customers about a connection between the two businesses, and that you’ve suffered or are likely to suffer actual damage — a general commercial-impersonation principle, not an AI-specific rule, but one that doesn’t care what tool produced the confusing content.
Don’t assume a disclaimer or a technically-true label makes a commercial deepfake safe — the deceptive-marketing and passing-off analyses both turn on the effect on the audience, not on how the content was labelled internally. And don’t assume the opposite either: making synthetic media of a real person isn’t a crime by itself under any Canadian statute found here.
The related detection question — whether you can reliably detect AI-written text — and the labelling question, whether AI content needs a label, both feed into how exposed a business is here. Diligence on a target using synthetic media commercially is exactly the kind of question Treadstone’s AI Due Diligence work is built to catch.
See how synthetic-media and provenance risk gets checked before closing.
