Canada Launches a National AI Council as the FTC's AI Accuracy Proposal Waits

Two developments on either side of the US–Canada border matter for any firm that sells or relies on AI tools in North America. One is new this month; the other is a US proposal from the summer that is still worth understanding.

Canada launches the Prime Minister's National Council on Artificial Intelligence

On 2 October 2026, Prime Minister Mark Carney announced the launch of the Prime Minister's National Council on Artificial Intelligence. The council will advise on the evolution of the government's "AI for All" strategy, which the government describes as its new national strategy for adopting AI responsibly — building trust, expanding opportunity and strengthening Canada's sovereign control over critical AI capabilities.

The council's members include Yoshua Bengio, Gillian Hadfield, Valérie Pisano and Michelle Zatlyn, alongside researchers, investors and former public servants. Its role is advisory: it will advise on accelerating adoption, making AI safer and protecting democracy. It does not make law.

That distinction matters. Canada still has no federal AI statute. The proposed Artificial Intelligence and Data Act died in January 2025, when Parliament was prorogued. For businesses using AI in Canada, the rules that bite today are privacy law — PIPEDA federally, and provincial law such as Quebec's Law 25 — and Ontario's rules on disclosing AI in job postings. A national strategy and an advisory council signal where policy may go; they do not change what your firm must do this week.

The FTC's proposed policy on AI accuracy

In the United States, the Federal Trade Commission proposed a policy statement in July 2026 on how Section 5 of the FTC Act — the ban on deceptive acts or practices — applies to companies that market AI systems. It was published in the Federal Register on 7 July 2026, with comments due by 31 July 2026.

The proposal targets AI systems configured to distort their outputs towards undisclosed objectives, contrary to what users reasonably expect. It also states that ordinary errors — "hallucinations" that come from the technology's limits rather than a deliberate design choice — do not by themselves raise Section 5 concerns. It is a proposed statement, not a final rule.

For a professional services firm, the practical lesson is simpler than the politics around it: what you tell clients about an AI tool must match what the tool actually does. If you resell, configure or recommend AI tools, keep a written record of what each one is for, what it is not for and who checks its output.

What to do now

• Know which rules actually apply. In Canada that means privacy law and, in Ontario, job-posting disclosure — not a federal AI act that does not exist.
• Check your AI claims. Anything you say about an AI tool in marketing or proposals should be something you can show is true.
• Keep a human in the loop. Every AI output that reaches a client, a court or a regulator should be checked by a named person first.
• Write it down. A short AI policy, an inventory of the tools in use and a named owner for each is the record a regulator or client will ask for.

Ops Intel writes AI compliance frameworks for small and medium businesses in the US, Canada and worldwide. See how at opsintel.io.