On 5 October 2026, according to MLex, the Competition Bureau filed its response to Innovation, Science and Economic Development Canada's (ISED) consultation on AI transparency. The Bureau made three recommendations. Transparency should support well-functioning competitive markets. Obligations should be proportionate. Regulation should target harms rather than specific technologies. It also warned that "prescriptive, costly, or technically complex requirements may disproportionately burden smaller firms and new entrants."
This is a competition regulator telling a policy ministry that disclosure mandates have market-structure consequences. That is a useful intervention, and the ministry should treat it as more than a procedural comment.
The case for transparency rules
The case for mandatory transparency is serious. ISED launched the consultation on 23 July 2026, and its announcement lists five areas: identifying AI-generated content, helping people recognise when they are dealing with an AI system, consistent information about how systems are developed and what they can do, tracking serious AI incidents, and monitoring the activities of AI agents. Minister Evan Solomon said Canadians "need to know when they are interacting with AI systems." Users cannot judge synthetic media, chatbots or autonomous agents if they cannot tell what they are dealing with. Opaque systems can also hide discrimination, fraud and safety failures from the people harmed by them. Voluntary disclosure has a poor record when it conflicts with commercial incentives.
The consultation also does not commit to legislation. According to Miller Thomson's summary, the options run from legislation and regulation to guidance, voluntary codes, technical standards and procurement requirements. Submissions closed on 23 September 2026, and ISED plans a "What We Heard" report. The policy design is still open, which is why the Bureau's timing matters.
Why design matters more than the goal
The goal of transparency is sound. The risk lies in how a rule is written. A watermarking or provenance mandate, an incident-reporting regime and an agent-logging requirement all carry fixed compliance costs. A large incumbent spreads those costs across a huge user base and employs lawyers and engineers dedicated to compliance. A start-up with ten staff does not. The Bureau's phrase about "smaller firms and new entrants" describes a familiar pattern: compliance duties can become a moat for the firms that are supposed to be constrained.
The Bureau did not arrive at this view suddenly. Its earlier report on its AI consultation, summarised by Baker McKenzie, acknowledged calls for transparency rules to prevent the extension of market power. It also said such rules should be developed strategically so they do not erode competitive benefits. It noted that a small number of companies hold significant shares of certain AI input markets and that larger firms often control data, compute and expertise. The same report recorded that some submissions pointed to open-source AI and public cloud infrastructure as lowering barriers for smaller firms. Both claims can be true at once, and good transparency design should preserve the second without worsening the first.
What proportionate design looks like
The Bureau's "target harms, not technologies" principle points to a few concrete choices for ISED:
- Tier obligations by risk and reach. Serious-incident reporting, for example, belongs with systems that can cause serious harm. It should not apply to every model deployed by a small business.
- Prefer interoperable standards over bespoke formats. A single provenance or labelling standard that a small firm can adopt through open tooling costs far less than a patchwork of requirements.
- Start with guidance and codes. Mandates can follow where voluntary measures demonstrably fail. That sequence keeps the option of regulation without locking in technical requirements that age quickly.
- Test for competitive effects. Where a rule is proposed, ISED should ask whether it raises costs more for entrants than for incumbents.
The Google case shows what is at stake
The Bureau is not abstract about market power. It is pursuing an abuse-of-dominance case against Google before the Competition Tribunal (CT-2024-010). It filed the application on 28 November 2024. It alleges Google is the largest provider across the ad tech stack for web advertising in Canada and has tied its tools together, given its own tools preferential access to ad inventory, taken negative margins in some circumstances to disadvantage rivals, and dictated terms on which publishers may use rival tools. It seeks an order requiring Google to sell two of its ad tech tools, a penalty and a prohibition on the conduct. Google denies the allegations, and the Tribunal will decide.
The two files point the same way. A competition authority that brings a structural case against an incumbent has a credible interest in not helping that incumbent through other channels. That is not an argument for exempting big firms from transparency duties. It is an argument for writing duties that scale. A rule that Google can satisfy with a compliance team and a start-up cannot satisfy works against the competitive outcomes the Bureau is litigating for.
The case itself should also be judged on its merits and evidence. Structural remedies are a heavy intervention, and a pro-innovation reading requires that they rest on proven conduct, not on a firm's size. The Tribunal process is the right place for that test.
What to watch
The next signal is ISED's "What We Heard" report. If it carries the Bureau's framing, with risk-tiered duties, harm-based targets and an explicit competition test, Canada could build transparency rules that protect users without raising entry barriers. If it moves quickly to technology-specific mandates, the Bureau's warning will have gone unheeded. The Bureau has given the ministry a workable test, and the burden of showing that a given rule passes it should sit with the regulator proposing it.