Canada Wants Your Say on What AI Must Disclose
Matthew Mbaka · September 15, 2026 · Canada Tech
Canada is asking a fairly practical question: when AI affects you, what should you be told?
The federal government's public consultation on AI transparency is open until September 23, 2026. Anyone can complete the anonymous survey or send comments by email. Submissions may be posted publicly, so the government says not to include personal or confidential information.
This is a consultation, not a new rule. Innovation, Science and Economic Development Canada says the outcome has not been decided. After submissions are reviewed, the department plans to publish a What We Heard report.
The five questions behind the consultation
The discussion paper covers five areas that show up in everyday technology:
- AI-generated content: Whether images, audio, video or text should carry reliable signals showing how they were made.
- AI interactions: When a person should be told that a chatbot or another automated system is speaking to them.
- Information about AI systems: What developers and businesses should disclose about capabilities, limits, training and appropriate uses.
- Serious incidents: Who should report an AI failure, when it becomes serious enough to report, and whether reports should be public or confidential.
- AI agents: How people can see what an agent was allowed to do, what it actually did and whether a person remained in control.
The government's discussion paper also puts the trade-offs on the table. A notice can help, but constant notices can turn into banner blindness. Public incident reporting can reveal dangerous patterns, but a badly designed rule may discourage companies from reporting problems or expose sensitive security details before a fix is ready.
A label should tell you something useful
A generic label that says "AI was used" may not help much. A photo editor can remove dust from a scanned family picture, while another system can invent a convincing event that never happened. Both may involve AI, but the risk is not the same.
A useful response to the consultation can name the decision a person needs to make. For example:
- A voter may need to know whether realistic political audio was synthetically generated.
- A job applicant may need to know whether an automated system screened or ranked the application and how to challenge an error.
- A customer may need to know when a support agent is automated, especially before sharing financial or health information.
- A business buyer may need clear retention, security and failure information before connecting a model to company data.
That level of detail is more useful than simply asking for "more transparency." It tells policymakers who needs the information, at what moment and for what decision.
AI agents need an activity record, not just a badge
The agent section may be the most consequential part of the paper. A small chatbot badge does not explain which mailbox an agent read, whether it placed an order or who approved a refund.
For an agent that can act, useful transparency could include a plain-language permission list, a timestamped activity history, confirmation before high-impact actions and a clear way to reverse them. Organizations may also need a chain of custody when one agent hands work to another.
ISED says 19.2 per cent of Canadian companies reported using AI to produce goods or deliver services in the second quarter of 2026, up from 12.2 per cent a year earlier. That makes these details less theoretical. They affect software purchases, workplace decisions and customer service now.
How to send a useful response
The consultation page links to the anonymous survey and also accepts emailed comments. A short response can still be useful if it gives one concrete example, explains the harm or confusion, names the disclosure that would have helped and says who should provide it.
Responses are due September 23. Because submissions may become public, remove names, account details, confidential workplace material and anything else you would not want posted online.
Tags: Policy, AI Regulation, Consultation, Canada