Filed with Innovation, Science and Economic Development Canada · August 2, 2026
In July, Ottawa asked how AI systems should be made more transparent. I submitted my response on August 2. The consultation closed on September 23, and the government’s summary of what it heard is expected in the coming months.
Many of the published responses come from universities or law firms. My view comes from the field. I advise the vendors selling AI into the Canadian mid-market, the channel partners deploying it, and the organizations deciding whether to buy. That table is where transparency policy actually succeeds or fails.
The submission offers thirteen recommendations. Five practical realities sit behind them.
We need to target what a system does, not just whether AI is under the hood. A deepfake video is dangerous because it is false. A proposal drafted with AI and checked by an expert is just ordinary work. If we force companies to flag every use of AI, we add cost without adding safety, and we teach the market that using AI is something to apologize for.
About one in five Canadian businesses used AI in the second quarter of 2026. Almost none have a Chief AI Officer or a data science team. General managers are making these calls based on vendor slide decks. Giving them dense technical documentation does not help. They need clear tools to figure out if their own company is actually ready to deploy these systems safely.
They buy from resellers, integrators, and managed service providers. A rulebook written only for developers and end users targets two groups who rarely speak to each other. Obligations have to travel down the commercial chain.
Information is only useful if it arrives before a mistake is made. If an AI system makes a claim, the user needs to see the source material right then and there. Knowing a system uses AI does not stop it from inventing fake legal citations, but showing the underlying source would.
Safe Harbor, Privacy Shield, and TRUSTe all failed in well-documented ways. As Canada designs its Trusted AI Certification program, we need to build it on independent verification, not self-attestation. And it cannot be so expensive that it locks out smaller Canadian innovators.