Policies Public-facing policies

AI Human Oversight Policy

Which AI-involved decisions must route through a human reviewer before becoming final.

Version
0.1.0
Effective
May 20, 2026
Last reviewed
May 20, 2026
Review cycle
every 6 months
Master Policy Index entry
§4 #31

Policy text

The AI Use Policy says people remain responsible. This policy operationalizes that promise: it names the decision categories where AI cannot be the final authority — suspensions, ownership disputes, civic escalations, accusations, child-safety issues, law enforcement responses, high-impact municipal reports, Indigenous data decisions, public allegations, and the publication of any sensitive location.

It applies to Map.ca product, AI, moderation, and operations teams. Every decision in those categories carries a human-reviewer audit record.

Principles this policy enforces

  • AI assists, people remain responsible
  • Civic reporting must be useful without becoming a public shaming system
  • Public data and personal data are not the same thing
  • Community benefit must survive scale

What it requires

  • Route every covered decision through a human reviewer before it becomes final.
  • Record the reviewer’s name and the decision in the audit log.

What it forbids

  • Do not allow AI to take irreversible action on covered decision categories.
  • Do not let throughput pressure justify skipping human review.

How it applies

References