Deploying AI at Protected B means the platform — not just the data — must live inside the control envelope: Canadian infrastructure, PBMM-aligned controls, auditable access, and no silent egress to foreign services. MAIA is architected for this from the substrate up: Canadian residency by default, deployable in a department's own cloud tenancy or on-premise, with every AI decision sealed into a tamper-evident audit ledger.
The envelope, not the checkbox
Protected B is a data classification, but in practice it is an architecture test. A platform that stores records in Canada while routing inference, telemetry, or support sessions through foreign endpoints has not met the intent of the control set. The honest assessment starts with the full dependency chain: where does every byte go, including the ones the vendor considers operational exhaust?
MAIA's answer is structural. The substrate runs where the institution runs — Canadian cloud regions, a departmental tenancy, or disconnected environments — and its dependency chain is disclosed and contractable. Nothing in the decision loop requires a call home.
PBMM alignment as a working posture
The platform's control architecture is mapped to the PBMM profile (Protected B, Medium Integrity, Medium Availability): role-based access on every surface, security-event logging, encryption in transit and at rest, and human authorisation gates on consequential actions. Assessment artefacts are available under NDA through our Trust Center.
Because every AI-drafted decision carries its evidence — signals read, policy applied, human approver — the audit story is not a separate compliance layer. It is the same ledger the operators use.