Users are entitled to know when they are looking at, or interacting with, AI. The AI Transparency Label Policy defines the surfaces that require AI labelling (AI-generated text, AI summaries, AI-suggested tags, AI-translated content, AI-assisted moderation states), the label format, the placement rules, and the accessibility requirements for the label itself. The shared label component is rendered on every such surface — the MapAi guide, pin classification and suggested tags, ChatMap suggested replies, and any other AI-produced output — so the disclosure is consistent and never hidden behind a hover-only affordance.
This policy also discloses the model behind the label. Map.ca runs a hybrid hosting model: hard reasoning may use a frontier model, but any step that touches personal information runs on self-hosted open-weights models inside a Canadian VPC. No personal information is sent to third-party models. User content is never used to train third-party models without explicit, separate opt-in — the default is no. Prompts and traces are minimized and redacted, consistent with PIPEDA and Quebec Law 25 and with the AI Use Policy.
The plain-language version of this disclosure lives at the public AI transparency page (/ai), linked from the AI label itself and from the privacy policy, so a user can move from any AI surface to a full explanation of how the AI works and where their data lives.
It applies to the Map.ca product teams, the AI team, and the design team responsible for the label component.
Requirements
- Label every AI-generated or AI-assisted user-visible output with the shared label component.
- Make the label keyboard- and screen-reader-accessible, and link it to the AI transparency disclosure.
- Disclose the hybrid hosting and data-residency model on the public AI transparency page.
- Keep personal-information-touching AI steps on Canadian-VPC open-weights models; never send personal information to third-party models.
Prohibitions
- Do not hide the AI label in a tooltip or mouse-only affordance.
- Do not let label visibility depend on a marketing decision.
- Do not send personal information to third-party AI models.
- Do not train third-party AI models on user content without explicit opt-in.