Pre-completed PIA template aligned with the TBS Directive on Privacy Impact Assessment for Canadian federal government departments deploying AccessPoint

Last updated: August 06, 2026 by Steve

GC Privacy Impact Assessment

This page is a public summary of the Privacy Impact Assessment (PIA) prepared by the software publisher (Realizer Services Inc.) for AccessPoint, aligned with the Treasury Board of Canada Secretariat (TBS) Directive on Privacy Impact Assessment. It documents the personal information holdings, information flows, privacy risks, and safeguards built into the product.

The publisher's PIA covers the technical and architectural sections and is intended to help Canadian federal institutions complete their own PIA. The deploying institution completes the institutional details (institutional overview, retention schedules, information sharing agreements, and sign-off) and submits the completed PIA to TBS and the Office of the Privacy Commissioner (OPC). The full PIA is available to customers on request — see the closing note.

For the full technical architecture, see the Technical Architecture page. For ITSG-33 security control mapping, see the GC Security Controls Reference page.

Purpose and Scope

AccessPoint is an access-to-information and privacy management platform that helps government institutions administer access and privacy programs in compliance with the Access to Information Act (ATIA), the Privacy Act, and equivalent provincial/international regimes. Its primary function is subject access / access-to-information request management — the full lifecycle from intake through assignment, document collection, review, redaction, and response packaging.

Beyond request handling, AccessPoint also supports the surrounding privacy program: Privacy Impact Assessments (PIA/AIA/Security assessments), privacy incident and breach management (including breach-notification workflows), complaints and appeals, a privacy risk register and commitments register, and records of processing (ROPA). These modules process additional categories of personal information and are covered by the same safeguards, roles, audit ledger, and data-residency guarantees as the request-management core.

Business process: intake (an SAO creates the request) → assignment (custodians receive sanitized instructions without requestor PII) → document collection → review and redaction → response packaging → closure (the retention clock begins).

Individuals affected: requestors; institution employees (SAOs, Administrators) whose actions are recorded in the audit trail; custodians and contributors (who do not have access to requestor PII); third parties whose PI may appear in documents under review; data subjects described at the category level in ROPA and assessments; individuals affected by a privacy incident; complainants and complaint parties; and consultation contacts.

Legal authority: collection and use rest primarily on the ATIA (ss.4, 6, 9, 19) and the Privacy Act (ss.4, 5, 7, 8(2)(m), and the individual's right of access).

Personal Information Inventory

Personal Information Collected and Processed

Data Element Sensitivity Source Purpose Stored In
Requestor full name Medium Requestor (via intake) Identify requestor, correspondence Azure SQL
Requestor email address Medium Requestor Email correspondence Azure SQL
Requestor mailing address Medium Requestor Mail response delivery Azure SQL
Requestor phone number Low-Medium Requestor Telephone correspondence Azure SQL
Requestor organization Low Requestor Statistical reporting Azure SQL
Request description Medium Requestor Define scope of search Azure SQL
Subject name Medium-High Requestor (via intake) Identify the subject of the request (may differ from requestor) Azure SQL
Subject date of birth High Requestor (via intake) Identity verification for privacy requests Azure SQL
Employee names, emails, Entra object IDs Low Entra ID User identification, notifications, authentication, audit Azure SQL
Audit trail (user actions) Low-Medium System-generated Accountability, compliance Azure SQL
Documents under review Potentially High Institution records ATIA response preparation Azure Blob Storage
Redacted document versions Medium System-generated Response packaging Azure Blob Storage
Attestation records and e-signatures (typed name) Low-Medium Custodians Formal sign-off on completeness Azure SQL
IP addresses and user agent strings (attestation, document access) Low-Medium System-captured Forensic audit of signing/access context Azure SQL
Requestor communications Medium SAO staff Correspondence records (direction, method, date, notes) Azure SQL
Delegation records Low-Medium SAO/Administrator Officer responsibility delegation Azure SQL
Consultation records and contact directory Low-Medium SAO staff Inter-departmental or external consultation Azure SQL
Requestor contact directory Medium Requestor / intake staff Reusable requestor identity; the request links to a contact rather than storing PII inline Azure SQL
Requestor conduct log Medium ATIP staff / system-generated Append-only, dated, attributed conduct observations and review outcomes — the evidentiary record behind a frivolous/vexatious determination. Entries are never edited; deletion is Administrator-only and audited Azure SQL
Requestor conduct analysis memo Medium ATIP staff (optionally AI-drafted, human-reviewed) Written analysis supporting or declining an F&V determination; can be rendered as a formal-record PDF Azure SQL / Azure Blob Storage
Vendor register contacts Low-Medium Privacy staff Business contact details of third-party processors/vendors, alongside DPA/contract facts and review cadence Azure SQL
Incident affected-PI details Potentially High Privacy/ATIP staff Incident facts, affected-PI categories, harm assessment, breach-notification tracking Azure SQL
Complaint parties and correspondence Medium Privacy/ATIP staff Complainant/party identity and complaint handling Azure SQL
Records-of-processing (ROPA) categories Low-Medium Privacy staff GDPR Article 30 descriptions by category of data subject/recipient — not individual records Azure SQL
Assessment questionnaire answers Low-Medium Privacy staff / delegates PIA/AIA/Security content; may describe PI handled by a program Azure SQL
Microsoft 365 captured records (opt-in) Potentially High The custodian's own M365 data Custodian-captured email, calendar, OneNote, Teams chats, or Copilot interactions as a responsive record Azure Blob Storage (converted PDFs)
Extracted document text Potentially High (mirrors document content) System-generated (incl. optional OCR of image-only scans) Plain text extracted for content search, duplicate detection, redaction suggestions, and AI grounding Azure SQL; optionally indexed in the institution's own Azure AI Search resource
AI document summaries (opt-in) Medium (may describe PI in the document) System-generated (Azure OpenAI in the institution's subscription) Cached machine-generated summary per document and language, regenerated on content change Azure SQL
AI usage metadata (opt-in) Low System-generated Metadata-only log of AI Assist calls: feature, model, token counts, calling user, and human accept/dismiss decisions. Prompt and response bodies are never stored Azure SQL
Hash-chained audit ledger Low-Medium System-generated Tamper-evident, append-only ledger of actions across all modules Azure SQL

Personal Information NOT Collected

AccessPoint does not collect or process: Social Insurance Numbers; financial information (bank accounts, credit cards); health or medical records as structured data (they may appear in documents under review); biometric data; criminal record information; location data; or cookies/tracking identifiers.

Note: IP addresses and user agent strings are captured only in limited contexts (attestation signing and document access logging) for forensic audit purposes, as listed above. General browsing behavior and device fingerprinting are not collected.

Sensitive Personal Information in Documents

Documents collected during the ATIA process may contain any category of personal information, including sensitive PI about third parties. AccessPoint stores, previews, and facilitates redaction of documents, and — to support the review tools — extracts each document's plain text into the database (with optional OCR of image-only scans via an in-subscription Azure AI Document Intelligence resource). That extracted text powers content search (optionally indexed into the institution's own Azure AI Search resource), duplicate detection, pattern/rule-based redaction suggestions, and (when the optional AI Assist component is deployed) machine-generated suggestions and summaries. All of this processing occurs within the institution's own Azure subscription, and every machine-produced redaction or suggestion is a proposal: classification and exemption of PI within documents is decided by trained SAOs using the built-in review tools.

Microsoft 365 record capture (opt-in, feature-toggle-gated): a custodian can capture their own Microsoft 365 records — Outlook email, Outlook calendar, OneNote, Teams chats, Microsoft Lists, or Copilot interaction history — as a responsive record. Capture always uses the signed-in user's own delegated identity and is scoped to that user's own data (Copilot retrieval is app-only, as no delegated Graph permission exists, but is always scoped server-side to the signed-in user's own Entra object ID — never tenant-wide). Captured content is rendered to a PDF and enters the normal review/redaction workflow. Calendar capture excludes items marked Private/Personal/Confidential by default.

Personal Information Flow Analysis

Stage Summary
Collection PI is collected under ATIA s.6 (request submission requires name and address) and Privacy Act s.4 (collection related to an operating program). Only PI necessary to process the request is collected, directly from the requestor and entered by SAO staff. Fields are configurable by the institution.
Use Requestor PI is used solely to process the ATIA/Privacy Act request. Only roles granted the view-requestor-PII permission (typically the Administrator, SAO, and Reviewer archetypes) can view requestor PII — Custodians and Contributors never see it (enforced server-side by PII-filter middleware). AccessPoint makes no automated decisions about individuals: the optional AI Assist component produces only suggestions, editable drafts, and read-only answers, and every exemption, redaction, disclosure, and determination decision is made by trained staff. Fixed statistical reports are aggregate only; the custom Report Builder exposes record-level fields only to roles holding the reporting permission, visibly flags PII fields in its catalog, and never offers PII fields to the AI "Ask AccessPoint" feature.
Disclosure The response package is provided to the requestor (ATIA s.7). Internal disclosure to SAO/Reviewer staff is role-controlled; Custodians/Contributors receive sanitized instructions only. Email and Teams notifications to custodians/contributors carry no requestor PI. The publisher receives only the tenant ID, API version, and the API's own base URL (license validation and web-part discovery), notification metadata (recipient user ID, activity type, actor display name, request number, preview/topic text, related record reference), and jurisdiction-pack installation reports — no requestor PI. Microsoft processes data as a sub-processor within the institution's own tenant.
Retention and disposal A built-in Retention Review feature applies configurable retention periods per record type across five registers (requests, assessments, incidents, complaints, and standalone risks) through one RetentionService. Purge permanently deletes the record with its documents, assignments, tasks, attestations, and related history — re-evaluating eligibility at purge time and blocking records with cross-case dependencies — and is logged in the RetentionPurgeLog table (which is itself never purged). Azure SQL automated backups retain data for 7–35 days depending on tier (configurable long-term retention).
Accuracy Requestor PI is entered from the original request form onto the reusable requestor contact record (every request links to a contact); employee PI is sourced from Entra ID and refreshed on each sign-in. The system does not enrich records from external sources or profile individuals. It does produce derived representations of content already held: extracted document text (including optional OCR), machine translations of staff-entered narrative fields (never overwriting a human translation), and — when the optional AI Assist component is deployed — cached document summaries and suggestion/draft outputs that staff accept or dismiss.

The notification path illustrates the "no requestor PI to the publisher" control:

Event (e.g., assignment created)
        │
        ▼
  API Notification Dispatch Service
        │
        ├── PII Filter: strips requestor PI for Custodian/Contributor templates
        │
        ├──► Email: Graph Mail.Send → Institution's shared mailbox → Recipient
        │    (via managed identity or institution's app registration)
        │
        └──► Teams: API → Realizer Platform API → Graph TeamsActivity.Send → Recipient
             (sends: user ID, activity type, actor name, request number, preview text — NO requestor PI)

Privacy Risk Assessment

Risk Likelihood Impact Mitigation Residual Risk
Unauthorized access to requestor PII Low Medium Granular permission-based RBAC with server-side enforcement; Custodians/Contributors architecturally excluded from PII; Entra ID MFA; permission checks on every request. Low
PI breach via document exposure Low-Medium Medium-High Encryption at rest (TDE, AES-256); role-based authorization; no public document access; redaction tools remove PI before response packaging. Low-Medium
PI in notifications Very Low Low Custodian/Contributor templates auto-strip PII; SAO notifications include the request number but not requestor identity; Teams notifications carry only notification type, request number, and task/assignment name. Very Low
PI transmission to publisher Very Low Low The publisher receives only the tenant ID and API address (license validation) and notification metadata — no PI, documents, or request details — and has no standing access to the institution's Azure resources. Very Low
PI outside Canadian jurisdiction Config-dependent Medium Data is stored in the institution's Azure subscription in the region it selects (Canada Central/East recommended); the Microsoft Cloud Agreement addresses data residency. Institution to assess
Inadequate retention/disposal Low-Medium Medium Built-in Retention Review with configurable periods and purge logging; institution aligns schedules with Library and Archives Canada disposition authorities. Low (with proper configuration)
Insider threat (authorized-user misuse) Low Medium Comprehensive audit trail with user attribution and timestamps; role-based access limits exposure; Administrators control role assignments. Low

Risk Area Scores

Per the TBS standardized framework:

Risk Area Score Rationale
Type of program or activity 2 Administration of program/activity (ATIA administration)
Type of personal information 3 Contact information and potentially sensitive records in documents
Program partners 2 Microsoft (sub-processor); Realizer (no PI access)
Duration of the program 4 Long-term/ongoing statutory obligation
Program population 3 External individuals exercising statutory rights
Personal information transmission 2 Encrypted cloud infrastructure within Canadian jurisdiction
Technology and privacy 2 Standard web application — no surveillance, biometrics, or automated profiling. Optional AI Assist features (Azure OpenAI in the institution's own subscription) are suggestion/draft-only — a person makes every decision, and only usage metadata is logged. Institutions deploying the optional AI components should reassess this score.
Potential impact on the individual 2-3 Possible embarrassment or reputational harm if requestor identity disclosed

Overall risk level: Moderate — standard for an administrative program processing PI of external individuals, mitigated by strong technical controls (encryption, RBAC, PII filtering, data sovereignty).

Technical and Administrative Safeguards

Authentication and Access Control

Safeguard Implementation
Authentication Microsoft Entra ID with JWT bearer tokens. No local user accounts or passwords.
Multi-factor authentication Enforced by the institution's Entra ID Conditional Access policies.
Continuous Access Evaluation Supported — token validation occurs on every API request.
Role-based access control Granular, tenant-configurable permission roles enforced server-side on all API endpoints. Administrator is the only built-in role; tenants define roles matching the archetypes (SAO, Reviewer, Custodian, Contributor, Reader), typically seeded from a jurisdiction pack.
PII filtering Server-side middleware strips requestor PI from responses for any caller not holding the view-requestor-PII permission; it cannot be bypassed by the client.
Session management Token-based; lifetime governed by Entra ID policies.
First-user bootstrap The first user is granted the Administrator role. No default or shared credentials exist, and a server-side guard prevents removing the last active Administrator.

Encryption

Layer Implementation
In transit TLS 1.3 minimum at the App Service front door; TLS 1.2+ on Azure SQL and Blob Storage; FTPS disabled.
At rest — database Azure SQL Transparent Data Encryption (TDE) with Microsoft-managed keys.
At rest — documents Azure Blob Storage AES-256 encryption; customer-managed keys (CMK) available.
At rest — secrets Azure Key Vault, accessed via managed identity.

Network Security

Safeguard Implementation
HTTPS only All HTTP traffic rejected; HTTP/2 enabled.
CORS Restricted to the institution's SharePoint domain.
Rate limiting 600 requests/minute per authenticated user.
Security headers X-Content-Type-Options, X-Frame-Options, Referrer-Policy, Permissions-Policy, Content-Security-Policy.
Management plane SCM/FTP basic authentication disabled; management via Azure Portal with Entra ID + MFA only.

Monitoring and Audit

Safeguard Implementation
Application audit trail All create/update/delete actions recorded with user ID, timestamp, field changed, and old/new values.
Tamper-evident audit ledger A hash-chained (SHA-256), append-only ledger records actions across all modules; integrity is verifiable server-side, and a court-ready case audit export (formerly "evidence package") can be produced for a request.
Document access log Preview and download events recorded with user ID, timestamp, IP address, and user agent. Insert-only.
Retention purge log All purge operations recorded with user attribution.
Application monitoring Azure Application Insights captures HTTP requests, exceptions, and dependency calls (90-day retention).
Alert rules Configurable metric alerts for HTTP 5xx errors and high latency.
Log Analytics Centralized log storage with KQL query capability for security investigations.

Administrative Safeguards

Administrators assign roles via Settings; role changes take effect immediately. An Application Access Policy restricts Mail.Send to the designated shared mailbox only. Configuration changes are applied via versioned jurisdiction packs with install audit logging. Staff training, acceptable-use policy, and breach-response procedures are specified by the institution.

Third-Party Services and Data Sharing

Service Role PI Received Data Location
Microsoft Azure Sub-processor (IaaS: App Service, SQL, Blob, Key Vault, Application Insights) All application data Institution's Azure region
Microsoft 365 Processor (Graph Mail.Send, Teams activity feed) and, optionally, a source of custodian-captured records Email/Teams notification content; for opt-in capture, that custodian's own M365 content Institution's M365 tenant
Realizer Services (Publisher) Software publisher — no processor role for customer PI None — tenant ID + API base URL + notification metadata + jurisdiction-pack install reports only Publisher's Azure (Canada)
Syncfusion Embedded library (not a service) None — conversion occurs within the institution's App Service Institution's App Service
Azure AI Search (opt-in) Institution-deployed resource — content search index Document metadata, extracted text, and content embeddings Institution's Azure subscription
Azure OpenAI (opt-in — AI Assist) Institution-deployed resource — suggestions, drafts, summaries, translation Bounded excerpts of case records and extracted document text at call time Institution's Azure subscription
Azure AI Document Intelligence (opt-in — OCR) Institution-deployed resource — text from scanned documents Document pages sent for prebuilt-read analysis Institution's Azure subscription

Optional Azure AI services. The three AI resources above are not publisher services: each deploys into the institution's own subscription and region, authenticates with the App Service managed identity only (API-key access disabled), and sends nothing to Realizer Services. Every AI output is a suggestion, an editable draft, or a read-only answer — staff decide; per-feature tenant toggles and a monthly token budget (which hard-stops all AI processing) provide governance; Microsoft does not train models on the data; and AccessPoint stores only call metadata, never prompt or response bodies. Where Azure OpenAI inference is processed depends on the deployment type chosen at deploy time (global or EU/US data zone) — institutions should confirm the choice against their residency requirements. Case audit exports for a request include an AI-involvement disclosure (metadata only), and the same information is visible on each record's Activity feed.

Key point: No requestor PI, document content, or request details are transmitted to Realizer Services or any external party. All document conversion and redaction processing occurs within the institution's own App Service, and the optional search, OCR, and AI processing occurs within the institution's own Azure subscription.

Applicable Personal Information Banks

PIB Registration Number Description
Access to Information Act and Privacy Act Requests PSU 901 Records related to ATIP requests
Employee Personnel Records PSE 901 Employee names and roles in the audit trail

Institutions should confirm applicable PIBs and whether new PIBs are required for their specific deployment.

Availability of the Full PIA

The full Privacy Impact Assessment — including the complete personal information inventory, detailed collection/use/disclosure/retention/accuracy analysis, legal authorities, and data flow diagrams — is available to customers and prospective customers on request. Deploying institutions use it as the technical and architectural basis for their own PIA, completing the institutional overview, retention schedules, information sharing agreements, and sign-off before submitting to TBS and the Office of the Privacy Commissioner.