“AI washing” is the term regulators use when a business’s claims about its own artificial intelligence outrun what the product actually does. In Canada, that is not a new category of law — it is the same false-or-misleading-representation rule the Competition Act has always applied, aimed at a newer kind of claim.
Key takeaways
Regulators reached for the term “AI washing” by analogy to “greenwashing” — a business overstating an environmental credential to capture demand for it. AI washing is the same move applied to a hotter label: describing a product as AI-powered, or overstating what its AI actually accomplishes, because the word itself now sells.
There is no dedicated “AI washing” provision in the Competition Act. What exists is the general prohibition on a false or misleading representation, applied to this specific kind of overstatement the way it has always applied to any other exaggerated claim.
The Competition Bureau’s discussion paper on AI and competition addresses this directly, and is careful about attribution: “The United States Federal Trade Commission (FTC) outlines a number of ways businesses can avoid making false or misleading representations around their AI products or services. These include exaggerating what an AI product can do, falsely promising that an AI product performs a task better than a non-AI product, not performing risk analysis, and fabricating that a product uses AI (when it may not at all).” (Competition Bureau, Artificial intelligence and competition, discussion paper) That is US enforcement material appearing inside a Canadian regulator’s own paper — useful as a description of the failure pattern, but not itself a source of Canadian legal obligation.
Canada does not need to import the FTC’s framework to act on AI washing, because the Competition Act already covers both halves of it. Overstating what an AI feature can do is a performance claim under paragraph 74.01(1)(b), which requires “an adequate and proper test” behind it and puts the burden of proving that test existed on “the person making the representation.” (Competition Act, s.74.01(1)(b)) Claiming a product uses AI when it does not is a separate, plainer problem — a false or misleading representation under paragraph 74.01(1)(a), assessed by the general impression it creates rather than only its literal wording. (Competition Bureau, the general impression test)
AI washing is not only a consumer-marketing problem — it is exactly the kind of overstatement a buyer needs to test rather than accept during due diligence. As a sister firm’s own writing on buying an AI business puts it, buyers “price proprietary data and a defensible model far above the wrapper around someone else’s API” — which means a target’s own marketing claims about its AI are precisely the claims a diligence process should be pressure-testing, not repeating in the deal memo. (deavo.ai, Buying & Selling an AI Business in Canada)
AI washing does not require an outright false statement to create a problem. The Bureau’s own list of misleading patterns includes a message that is “literally true but misleading because it does not include or state essential information that would likely influence consumer behaviour.” (Competition Bureau, the general impression test) A product described as “AI-enhanced” because it uses a minor, incidental AI feature, while the marketing implies AI does the core work a person actually does by hand, fits this omission pattern without a single sentence in the ad being technically false.
Part of what makes an AI performance claim so easy to overstate is that Canada currently has no published, independent accuracy or performance benchmark for AI tools to be checked against — Statistics Canada’s own business-conditions survey measures adoption and application mix, not performance. That absence cuts one way only: it makes the reversed-onus rule in paragraph 74.01(1)(b) more important, not less, because a business cannot point to an accepted external standard its product met — it has to have run its own adequate and proper test and be able to produce it.
A business that has signed ISED’s Voluntary Code has taken on a specific transparency commitment for a public-facing system: to “publish information on capabilities and limitations of the system.” (ISED, Voluntary Code of Conduct) Marketing copy that overstates a capability beyond what that same business has published elsewhere as its own stated limitation is not just a code lapse — it is the business contradicting its own public disclosure, which is exactly the kind of inconsistency a general-impression analysis is built to catch.
A pitch deck describes a “proprietary AI fraud-detection engine,” but the feature is a fixed set of manually written rules with no learned model behind it at all. That is a textbook version of the FTC’s fourth category — fabricating that a product uses AI — and under Canadian law it is reachable directly as a false or misleading representation under paragraph 74.01(1)(a), assessed by the general impression the term “AI-powered” creates for a reasonable reader, not by whether the term technically appears somewhere in the product’s documentation.
Related: AI claims and the Competition Act, and AI-generated reviews and testimonials.
How a buyer tests these claims before closing a deal is covered on the AI due diligence hub.
The Act does not define AI, and neither does the Competition Bureau’s own discussion paper, which states there is “still no universal definition for AI.” The operative test is not a technical taxonomy — it is whether the overall impression the term creates for a reasonable reader is accurate, under the same general impression test that applies to every other representation.
The business, in every case governed by paragraph 74.01(1)(b). The proof that an adequate and proper test existed “lies on the person making the representation,” not on a regulator to disprove the claim after the fact.
Yes, if the overall impression it creates is misleading even though no individual sentence is false — the general impression test specifically covers a literally true statement that omits information a reasonable consumer would find material.
A short call is enough to walk through a specific claim — yours, or a target’s — against the rule above.
