On 23 September 2026 the Competition Bureau filed a submission to the federal government's consultation on AI transparency. Its central line is blunt: "there is no 'AI exception' to the law, and for good reason." The Bureau also argues that "regulation should focus on harms, not technologies." The submission is policy advice. As law firm BLG noted when it reported the filing on 15 October 2026, it "does not create new legal obligations." It is still a useful signal about how Canada's competition enforcer intends to treat AI-era business conduct, including by the largest technology firms.
The case for something new
The strongest argument for AI-specific transparency rules is an information gap. Consumers cannot inspect a model, and a business buying an AI tool often cannot either. Markets with that kind of opacity can reward whoever overstates their product. The discussion paper that Innovation, Science and Economic Development Canada put out asks whether new measures are needed in five areas: AI-generated content, AI interaction, information about AI systems, AI incidents and AI agents. The consultation ran from 23 July to 23 September 2026, according to BLG. Those are real questions, and a regulator that dismissed them would be ignoring a genuine problem.
What the Bureau actually said
The Bureau's answer is that the law already reaches the conduct that matters. Its submission notes that "many laws, including the Competition Act, are technology neutral," and that frameworks focused on "the underlying conduct and its effects are generally more adaptable." Firms must comply whether they use AI, conventional software or something else.
It is worth being precise about the limits of the text. The submission itself does not discuss algorithmic pricing or vendor disclosure obligations, and it does not name big tech or dominant firms. Those applications come from BLG's reading. BLG says the Bureau's reasoning means businesses should check their AI marketing claims, their algorithmic pricing and their vendor contracts against the existing Act today. That is a fair inference. It is not something the Bureau has said in so many words.
The submission does make a competition argument that bears on large platforms. It warns that "prescriptive, costly, or technically complex requirements may disproportionately burden smaller firms and new entrants while favouring incumbents with greater compliance resources." It also says disclosure requirements should avoid "compelling the disclosure of information where doing so could undermine competition." A Canadian AI-labelling or audit regime would be cheap for a firm with a large legal department. A start-up would feel it far more.
The existing law has teeth
Part of the pro-innovation case is that the existing statute is not toothless. Section 74.01 of the Competition Act makes it reviewable conduct to make "a representation to the public that is false or misleading in a material respect" in order to promote a product or business interest. Section 74.01(1.1) separately treats a price that is "not attainable due to fixed obligatory charges or fees" as a false or misleading representation. That is the drip-pricing rule, and it applies whether the price came from a person or a model.
The penalties are large. Under section 74.1, a court can order a corporation to pay up to the greater of $10 million ($15 million for a subsequent order) and three times the benefit derived from the conduct. Where that benefit cannot be reasonably determined, the cap is 3% of annual worldwide gross revenues. A firm that markets a product as "AI-powered" when it is not, or presents a personalised price as a general one, faces that exposure today without any new statute.
Enforcement has been measured, and that is a feature
The Bureau's record on algorithmic pricing shows what conduct-based enforcement looks like in practice. It ran a public consultation on algorithmic pricing from 10 June to 4 August 2025. Its discussion paper said pricing algorithms "may facilitate both explicit and tacit forms of competitor collaborations, including 'hub-and-spoke' agreements." On 10 November 2025 it concluded a civil investigation into algorithmic pricing software in rental housing, naming RealPage Canada and Yardi Canada. It found adoption "did not meet the anticompetitive thresholds required to determine a violation" of the Act. It said it did not find sufficient evidence of abuse of dominance or anticompetitive collaboration "at this time." It also said it "remains concerned" about the effect on multi-family rental competition, and it issued guidance to landlords, property managers and software vendors.
Critics will read that as a regulator that looked and found nothing actionable. We read it as the right result. The Bureau investigated, applied an evidentiary threshold, and published guidance rather than inventing a violation because the software was novel. A technology-specific statute would have pushed it toward the opposite error, treating the tool as the harm.
Where the approach is weakest
The honest weakness of the technology-neutral position is speed and proof. Hub-and-spoke pricing through a shared vendor is hard to prove under a framework written for human conspirators. The Bureau's own paper concedes that algorithms can facilitate tacit coordination. If courts demand evidence of an agreement that a model made unnecessary, enforcement could lag the market. That is an argument for targeted amendments if specific gaps appear in litigation. It is not a case for a broad AI statute drafted before anyone knows what the gaps are.
What this means for large technology firms
The practical message is that Canada's AI compliance burden for big tech will, for now, come from laws already on the books. Firms should substantiate performance claims, audit how "AI" is used in advertising, and check that dynamic pricing does not present unattainable or personalised prices as general ones. BLG also suggests buyers negotiate contractual rights to documentation, testing information and incident notification from AI vendors. That is sensible risk management.
Ottawa should keep that restraint. A proportionate regime enforces fraud, deception and collusion as they occur and leaves the technology itself free to develop. If Parliament later legislates on AI transparency, the Bureau's warning about compliance costs falling hardest on entrants should be the test any bill has to pass. The submission does not settle the policy debate, but it gives that debate a useful standard: show the harm first, then regulate it.