Privacy Policy
Effective Date: November 14, 2026 — including the revisions of November 3 to 10, 2026 for AccessPoint 2.6.0 and Hosted AccessPoint (Section 15)
Applies to: this website, AccessPoint 2.6.0, Hosted AccessPoint, and Realizer's services around them
Publisher: Realizer Services Inc. | Contact: privacy@realizer.io
1. Introduction
This Privacy Policy describes how Realizer Services Inc. ("Realizer", "we", "us", "our") collects, uses, stores, and protects information in connection with AccessPoint, our access-to-information and privacy management solution for Microsoft 365 and Azure ("the App").
AccessPoint comes in two editions:
- The Microsoft 365 edition is designed with a data sovereignty model: customer data is stored and processed entirely within the customer's own Microsoft Azure subscription and Microsoft 365 tenant. Realizer does not host, store, or have standing access to that data.
- Hosted AccessPoint (
app.realizer.io, from version 2.6.0) is the same software run by Realizer: Realizer hosts the customer's data in its own Azure subscription in Canada and processes it on the customer's behalf (Section 5.9).
This policy covers these contexts:
- The Website — realizer.io and related pages
- The App — the AccessPoint software deployed in a customer's environment (the SharePoint web part, the Teams app, the API, the database, the storage account, and the optional Azure services the customer chooses to deploy alongside them)
- Hosted AccessPoint — the App as a service Realizer runs for the customer (Section 5.9)
- The Platform — Realizer's platform services that support licensing, Teams notifications, jurisdiction packs, first-time setup, Teams tab routing, and software distribution, and that register Hosted AccessPoint's organizations and keep their plans and billing
- Realizer's other services around AccessPoint — the AccessPoint Training Lab at
learn.realizer.io(Section 5.10), the marketing demo atdemo.realizer.io(Section 5.11), and Realizer's own customer records and admin console (Section 5.12)
2. Data Controller and Data Processor Roles
- Customer organizations are the data controller for all personal data processed within AccessPoint. They determine the purposes and means of processing personal data related to their access requests, privacy assessments, incidents, complaints, and risks.
- For the Microsoft 365 edition, Realizer Services Inc. acts as a data processor only to the extent it processes limited data through the Platform (see Section 5). Realizer does not control or determine the purposes for which customer data is processed within AccessPoint.
- For Hosted AccessPoint, Realizer is the customer's service provider — a data processor — for all the data the customer keeps in it (Section 5.9). The customer remains the controller; Realizer processes the data only to provide, secure and support the service.
- Training providers that run classes in the AccessPoint Training Lab are the controllers of their learners' data; Realizer is their processor under the provider agreement (Section 5.10).
- Realizer is the controller of its own business records: the account contact and its customer, billing and accounting records (Sections 5.8 and 5.12), the records it needs to run and secure its services, and the marketing demo's anonymous usage statistics (Section 5.11).
- Microsoft is the customer's cloud provider. Every Azure resource AccessPoint uses — including the optional AI, search, speech, and security services in Section 4.6 — is provisioned in the customer's own subscription under the customer's own agreement with Microsoft. Realizer is not a party to that processing. For Hosted AccessPoint, the Training Lab and the Platform, Microsoft is Realizer's subprocessor under Realizer's own agreement with Microsoft (Section 8.5).
3. Information Collected via the Website
3.1 Information You Provide
We collect information that you voluntarily provide, including:
- Contact form submissions: name, email address, company, and message content
- Demo requests: name, work email, organization, the jurisdiction or regime you work under, and anything you tell us you would like to see
- Where submissions go: contact and demo submissions are emailed to Realizer and, for demo requests, recorded in our customer relationship system (HubSpot, in Realizer's account in HubSpot's Canadian data centre in Montréal) together with the page you first arrived on, the site that referred you, and any campaign parameters in the link you followed, so we know how you found us
- The interactive demo: when you choose to book a demo from inside the interactive demo, its link back to this site carries the sample you tried, the jurisdiction you chose, how many steps you completed and any organization size you checked the price for; they are sent to us only if you submit the demo-request form, with the rest of your request
- Emailing yourself a tool's results: the self-checks and calculators in our guides can email you a link that reopens your results. For that we ask for your work email, your role and, if you choose, your organization's name. The email goes to you alone, once, through Azure Communication Services (Microsoft); your answers and score travel only inside that email, and we keep no copy of it. We do not store your address or add it to any list: the site keeps a one-way hash of it, in memory, for 24 hours, so that no address receives more than three of these emails a day, and its logs record only which tool and which role. We will not send you anything else unless a later version of this form offers it and you tick a box to agree.
- Service inquiries: information relevant to your onboarding or consulting needs
3.2 Automatically Collected Information
When you visit our website we use Google Analytics 4 to understand which pages are read and how visitors arrive. It records the pages you view and how long you stay, the site or search engine that referred you, your browser and device type, and an approximate location derived from your IP address; Google Analytics 4 does not log or store IP addresses. We have turned off Google signals and every advertising feature, so nothing is used for advertising and no profile is built across other websites. Visitors in the European Economic Area, the United Kingdom and Switzerland are asked before analytics cookies are set; elsewhere two first-party cookies (_ga and _ga_ followed by our property identifier) distinguish returning visitors. You can opt out on any device by opening any page of this site with ?ga=off added to the address, for example realizer.io/?ga=off, and ?ga=on turns it back on. The site sets no other tracking cookies.
Which button you pressed. Links from this site to the Hosted AccessPoint sign-up carry the name of the button pressed (cta, for example web-pricing), and links to the deployment page carry the name of the page they were on (via). These names describe our pages, not you; we count them in Google Analytics, and Section 5.9.6 describes what the sign-up keeps.
The interactive demo. The demo opens on its own page at demo.realizer.io. It uses the same Google Analytics property and the same consent choices as www.realizer.io, and counts only anonymous use of its screens — the sample, the language and the jurisdiction chosen, the steps completed and the buttons of its About panel. It never records anything typed in the demo, sets no advertising cookies and does not track you across other websites. The choice you make on its notice is kept in your browser (rz-consent), and ?ga=off opts you out on the demo too. Section 5.11 describes the demo in full.
4. Data Processed by the App (Customer Environment)
All of the following data is stored and processed exclusively within the customer's own Azure subscription and Microsoft 365 tenant. Realizer does not have access to this data.
For Hosted AccessPoint, the same descriptions apply to the data an organization keeps in it, with Section 5.9's differences: it is stored in Realizer's Azure subscription in Canada on the organization's behalf, and the Microsoft 365 parts (Section 4.5 and the Microsoft Graph permissions of Section 8.1) and repository connectors (Section 4.8) are not part of it.
4.1 User Identity Data
- Display name, email address, Entra ID object ID, and tenant ID
- Sourced from Microsoft Entra ID (Azure Active Directory) JWT tokens during authentication
- Used for authentication, authorization, audit logging, and people picker functionality
- Stored in the customer's Azure SQL database
- The web part keeps a 24-hour cache of the customer API's own address in the browser's local storage so it need not be rediscovered on every page load; the cache holds no user identifiers
4.2 Case Data
- Request details: titles, descriptions, request numbers, dates, statuses, priority levels, extensions, consultations, fees, hours, and correspondence logs
- Requestor information: names, contact details, addresses, organizational affiliations, preferred language, and requestor category — stored as reusable requestor contact records shared across a requestor's requests
- Requestor conduct records, where the customer chooses to record them: frivolous/vexatious determination details and notes, and an append-only conduct log attached to the requestor contact (removed when the contact is purged)
- Assignment and task records: custodian assignments, contributor tasks, instructions, due dates
- Attestation records: formal attestations with e-signatures
- Related case records, where the customer uses these modules: privacy incidents/breaches, complaints, privacy and algorithmic impact assessments, privacy audits, and risk register entries — including any personal data the customer records in them (e.g., individuals affected by an incident)
- Assessment validation records, where the customer uses them: the opinions asked for and given on an assessment (a data protection officer's or other advisor's advice, and the views of the people concerned — recorded as groups or representatives, never as named individuals), the organization's decision with the name and capacity of the staff member who took it, section evaluations, and consultations of a regulator (the authority, its reference, the dates and the outcome)
- Privacy audit contacts: the business contact details (name, organization, job title, email address, telephone) of the people an audit interviews, asks for evidence or reports to — shown only to the audit team and deleted with the audit
- Privacy program records: privacy subjects (programs, systems and activities) with their records of processing, which describe personal information by category rather than about individuals; the vendor and systems registers; program attestations; and discovery surveys (Section 4.11)
- Preservation holds, where the customer places them: the reason, the matter or file reference, and who placed and released each hold (Section 4.10)
- Audit history: timestamped records of all create, update, and delete actions with user attribution, backed by a tamper-evident hash-chained audit ledger
4.3 Documents and Recordings
- Files uploaded by users (Office documents, PDFs, images, email files, and any other file type the customer's records contain), including records users capture from their own Microsoft 365 content (see Section 4.5)
- Video and audio records: the original recording is never modified; AccessPoint derives a playback proxy, poster image, scrub thumbnails, waveform, transcript, captions, and — at release time — a burned release copy, all stored beside the original in the customer's Blob Storage
- Converted PDF versions for preview and redaction, and a reduced-size viewer copy for very large PDFs
- Redaction records: the geometry, exemption citations, status, reviewer, and — for suggested redactions — the text the detector matched. Detected and overlay text is subject to the same access gate as requestor PII
- Redacted document versions and response packages (ZIP and merged PDF exports), together with a manifest of file hashes
- Plain text extracted from converted documents (including OCR of scanned documents when the optional Document Intelligence resource is deployed), spoken content transcribed from recordings (when the optional Speech resource is deployed), and on-screen text read from video frames (when Document Intelligence is deployed) — stored in the customer's Azure SQL database to power content search, duplicate detection, AI summaries, and redaction suggestions
- Working documents (Word, Excel, diagram, and list files edited in place), with their version history
- The malware scan verdict for each file when Microsoft Defender for Storage is deployed (Section 4.6.5)
- Stored in the customer's Azure Blob Storage with Microsoft-managed encryption at rest (customer-managed keys optionally available) and shared-key access disabled — every reader authenticates with an Entra identity. A storage policy the customer switches on may later move a closed case's files to a cheaper access tier of the same storage account (Section 4.10)
Email conversion. When an email file is converted for preview, the customer's API fetches remote images the message references (at most 25 images of up to 4 MB each, over an SSRF-hardened client with an 8-second timeout) so the rendered PDF matches what the recipient saw, then renders the page offline so that nothing contacts the network during rendering or when the PDF is later viewed. Because the images are fetched, the image hosts named in the email — including a sender's "tracking pixel", if the message carries one — see a request from the customer's App Service at conversion time, which can tell a sender that the message was opened. No other party sees the message. Attachments are extracted into child documents by default (tenant toggle Features:ExtractEmailAttachments).
4.4 Configuration Data
- Tenant settings, notification templates, request types, choice fields, numbering configurations, jurisdiction pack content, custom fields, assessment templates, retention rules, and feature toggles
- Stored in the customer's Azure SQL database
4.5 Records Captured from Microsoft 365
The Documents workspace lets a signed-in user add records from their own Microsoft 365 content. Every capture is initiated by the user, gated by the tenant toggle Features:M365CaptureEnabled (with separate toggles for the request and assignment workspaces), and stored as an ordinary document in the customer's Blob Storage. Which identity performs the read matters for privacy, so each source is listed with its token:
| Source | What is captured | Token used | Notes |
|---|---|---|---|
| Outlook email ("My Email", shared/intake mailboxes) | Selected messages, downloaded as .eml files | The user's delegated token (Mail.Read, Mail.Read.Shared) | Exchange mailbox permissions still apply — a user sees only mailboxes they already hold access to |
| Outlook calendar | A date window of the user's own (or a shared) calendar, rendered as one day-grouped schedule record | The user's delegated token (Calendars.Read, Calendars.Read.Shared) | Items whose sensitivity is Private, Personal, or Confidential are excluded by default (opt-in). Attendees and location are included by default; descriptions and meeting join links are off by default. The record carries a banner noting the exclusion and that attendee names are third-party personal information subject to review |
| Teams chats and channel messages | The user's own chat and channel conversations, rendered as transcripts | The user's delegated token (Chat.Read, ChannelMessage.Read.All, Team.ReadBasic.All, Channel.ReadBasic.All) | Only conversations the user is a member of |
| OneNote | Pages from the user's notebooks | The user's delegated token (Notes.Read, Notes.Read.All) | |
| SharePoint lists and files | Lists rendered as records; files copied from sites and OneDrive | The user's delegated token (Sites.Read.All, Files.Read.All) | SharePoint permissions still apply |
| Teams meeting recordings and transcripts | Recordings of meetings the user organized, with the Teams-generated transcript | The user's delegated token (OnlineMeetingRecording.Read.All, OnlineMeetingTranscript.Read.All, OnlineMeetings.Read) | Graph lists recordings for the organizer only. The MP4 is downloaded through the user's browser into the upload queue and uploaded to the customer's API; the Teams WebVTT transcript is imported with it and is treated as authoritative — the recording is never re-transcribed. Nothing transits the publisher |
| Microsoft 365 Copilot interaction history | The user's own Copilot conversations, grouped by session, rendered as transcripts | The API's application permission (AiEnterpriseInteraction.Read.All) — there is no delegated permission for this data | Retrieval is performed server-side by the customer's API and is always scoped to the signed-in user's own Entra object ID, never tenant-wide. Requires the user to hold a Microsoft 365 Copilot licence. The permission is granted by default at setup with an opt-out |
Captured records are converted to PDF in the customer's App Service, stamped with their source (for example "Email Export", "Calendar Export", "Teams Meeting Recording", "Copilot Export"), and are indistinguishable from manual uploads thereafter.
4.6 Optional Azure Services
Customers may deploy the following services within their own Azure subscription by setting a deployment parameter. Each is provisioned by the same template as the rest of AccessPoint, in the customer's resource group, and each authenticates with the App Service's managed identity only — local authentication and API keys are disabled on every account that supports it (disableLocalAuth). None of them is required; an undeployed service simply leaves its feature unavailable. Document and case content processed by them does not leave the customer's Azure environment and is never sent to Realizer.
4.6.1 Azure OpenAI (AI Assist)
- Deployed by
deployAiAssist=true. Creates an Azure OpenAI account with a "chat" deployment (the everyday model), an "embed" deployment for vectors, and — unlessaiPremiumDrafting=false— a "chat-premium" deployment used for letters, narratives, and the recall-sensitive redaction pass. - Data processed: case fields, document text, transcripts, and the user's question, composed into a grounding prompt for the requested feature (document summaries, suggested redactions, drafting, intake triage, custodian suggestions, assessment prefill, Ask AccessPoint, case check, risk suggestions, auto-tagging, fee-category suggestion, machine translation of customer-entered text — including the questions the customer writes for a discovery survey — and vector embeddings for semantic search). AI features used on fictional training cases in train mode (Section 4.12) send that fictional content the same way and count against the same budget.
- Identity: the API's managed identity holds the Cognitive Services OpenAI User role; no key exists.
- Where processing runs: the account, its data at rest, and its endpoint stay in the region the customer chose (from version 2.6.0, where that region does not offer Realizer's recommended models and the customer named no AI region, a new account is placed in a region of the same geography). With the default Global standard deployment type, Microsoft may perform inference in other Microsoft regions; Data zone standard bounds inference to the EU or US data zone. The customer selects this at deployment (
aiDeploymentSku) and may pin the account's region (azureOpenAiLocation). Processing is subject to Microsoft's Azure OpenAI Service data, privacy, and security terms with the customer. - What is stored: the AI activity log (
AiAssistLog) records metadata only — feature, record identifier, model, prompt version, a hash of the prompt, token counts, duration, outcome, the initiating user, and the human accept/edit/dismiss decision. The composed prompt and the model's raw response are not persisted. Where a feature's output is something the user works with, it is stored as ordinary case data in the customer's SQL database: cached document summaries (per document and language), machine translations (stored per language and never overwriting a human translation), cached per-case artifacts (suggested next action, case check findings), suggested redactions (as proposals), and Ask AccessPoint conversation history (saved per user so a conversation can be resumed; a conversation about a case is deleted with that case, and one about no case is deleted after a period without activity the customer sets — 365 days by default; the optional question-mining feature summarizes themes across those conversations for administrators, and stops using a conversation as soon as it is deleted). - Human control: AI drafting outputs land in an editor for human review and are never sent automatically. AI-suggested redactions are recorded as proposals and are excluded from every export path until a human accepts them. Compliance-relevant findings in Case check are deterministic; the model may only add warnings and suggestions.
- Budget and opt-out: a monthly token budget (
Ai:MonthlyTokenBudget) stops all calls when exhausted. Every feature has its own toggle in Settings → Feature Toggles (Features:Ai*); the whole capability is removed by redeploying withdeployAiAssist=falseor deleting the account.
4.6.2 Azure AI Search (content and semantic search)
- Deployed by
deployAiSearch=true. - Data processed: an
accesspoint-documentsindex holding document metadata and the extracted text of each converted document (plus a vector when Azure OpenAI is also deployed), and anaccesspoint-casesindex holding vectors of request and assessment scope text for semantic duplicate detection, custodian suggestions, and prior-request retrieval. Results are always re-filtered against the caller's permitted document set, so a search can never surface a record the caller could not open. - Identity: the API's managed identity holds Search Index Data Contributor and Search Service Contributor; keys are disabled.
- Retention: index entries are written at upload and conversion, updated on re-conversion, and removed when the document or case is deleted or purged; orphaned entries are garbage-collected.
- Opt-out: without the resource, search falls back to metadata (name, type, tags). The semantic index additionally requires
Features:AiSemanticSearchEnabled.
4.6.3 Azure AI Document Intelligence (OCR)
- Deployed by
deployDocumentIntelligence=true. - Data processed: image-only PDFs (up to 300 pages per document) are sent page by page to the
prebuilt-readmodel and the recognized text is stored with the document; when the media features are used, sampled video frames (capped atRedaction:FrameScanMaxFrames, default 300) are read the same way so on-screen text can be searched and proposed for redaction. - Identity: the API's managed identity holds the Cognitive Services User role; keys are disabled.
- Region: Document Intelligence is hosted in a subset of Azure regions. When the customer's region does not offer it, the template places the account in the nearest supported region (for example Canada East → Canada Central) unless the customer overrides
documentIntelligenceLocation. Customers with strict regional requirements should confirm the resolved region. - Opt-out:
Features:OcrEnabled(documents) andFeatures:MediaFrameOcrEnabled(video frames) in Feature Toggles; or do not deploy the resource.
4.6.4 Azure AI Speech (transcription of recordings)
- Deployed by
deploySpeech=true. - Data processed: for each recording with an audio track, the API extracts a 16 kHz mono audio file into the customer's documents storage account and submits a batch transcription job. The Speech account reads that audio through its own system-assigned managed identity (Storage Blob Data Reader on the documents account) — no shared-access token is ever minted. When results are retrieved, the transcription job is deleted from the Speech service and the temporary audio file is deleted from storage. The transcript (with word timings), a caption file, and a plain-text rendering are stored with the recording.
- Identity: the API's managed identity holds Cognitive Services User and Cognitive Services Speech User; keys are disabled.
- Access to spoken content: since September 23, 2026 a transcript follows the recording it belongs to — anyone allowed to open the recording may read its words, because they can already hear them. A user without the requestor-PII permission gets the words with speaker names replaced by "Speaker 1", "Speaker 2" and so on, both in the transcript and in the captions track; the captions are unlocked by a separate token that never unlocks the video.
- Teams transcripts: a recording captured with its Teams transcript is never sent to the Speech service.
- Opt-out:
Features:MediaTranscriptionEnabledin Feature Toggles; or do not deploy the resource.
4.6.5 Microsoft Defender for Storage (malware scanning)
- Deployed by
deployDefenderForStorage=true. - Data processed: Microsoft Defender for Cloud scans every blob written to the customer's documents storage account on upload and records a verdict as a blob index tag. AccessPoint reads only that verdict and mirrors it to the document record; a file flagged as malicious can never be previewed, played, downloaded, included, or released. Defender's separate sensitive data discovery feature is explicitly left off — AccessPoint classifies content itself.
- Scope: enabled on the documents storage account only, capped at
malwareScanCapGBPerMonth(default 5,000 GB) of scanning per month. - Opt-out: do not deploy the module; the app setting
Documents__MalwareScanningis then absent and verdicts read as "not scanned".
4.6.6 Media processing worker (Azure Container Apps)
- Deployed by
deployMediaProcessing=true. Without it, recordings are transcoded and burned in-process on the customer's App Service. - Data processed: the worker reads the original recording from the customer's storage account, produces the proxy, poster, thumbnails, waveform, and release renditions, runs the face and on-screen-text sweeps, and writes the results back to the same storage account. It never connects to the customer's database or API — jobs and results travel through two Azure Storage Queues in the customer's subscription and carry blob paths, not content.
- Identity: the job's own managed identity holds Storage Blob Data Contributor on the
documentscontainer only and Storage Queue Data Contributor on its own two queues (media-jobs,media-results) only — nothing on the rest of the storage account (since version 2.4.0; the accessibility checker job,deployAccessibilityCheck, is scoped the same way todocumentsand itsaccessibility-jobs/accessibility-resultsqueues). The API's managed identity holds Storage Queue Data Contributor. The job scales to zero between bursts and each execution is a fresh process that ends by its replica timeout (mediaJobTimeoutHours, 8 hours by default). - Logging: the container environment writes its logs to the customer's own Log Analytics workspace.
- Container image: the worker is code, distributed like the API package. By default the job pulls
realizer.azurecr.io/accesspoint/media-worker:<version>anonymously from Realizer's container registry — see Section 5.6 for what the registry observes. For production, customers are recommended to mirror the image into their own Azure Container Registry (mediaWorkerImage+mediaWorkerRegistryResourceId), after which the job pulls with its managed identity and has no runtime dependency on Realizer. - In-tenant visual analysis: when a reviewer asks for it, AccessPoint detects faces in sampled video frames and tracks a region a reviewer has drawn across the clip so the same blur or box follows it. Both run inside the customer's subscription — in the media worker when it is deployed, otherwise in the customer's App Service — with an open-source detector (YuNet) whose weights ship inside the application; no cloud vision service is called and nothing leaves the tenant. This is detection only: each detection keeps a box, a time and a confidence score, and becomes a proposed redaction for a person to accept; no facial template, embedding, identity, or match is computed or stored. Face detection is on by default wherever recordings are used; a tenant switches it off with
Features:MediaFaceDetectionEnabled. - Direct playback (optional): when the template is given the tenant's SharePoint origin (
sharePointOrigin), the browser streams recordings directly from the customer's storage account using a one-hour user-delegation token instead of proxying through the API.
4.6.7 Document worker (Azure Container Apps, deployed by default)
- Deployed by
deployDocumentWorker=true(the default). A Container App with no inbound network access that runs an Office converter (Word, Excel and PowerPoint to PDF) and an HTML renderer (HTML and email to PDF) so neither runs on the App Service. - Data processed: the API places each file to convert in a
document-conversionscontainer and a job on thedocument-jobsqueue; the worker writes the PDF back and the API deletes the job files at once (a one-day lifecycle rule is the backstop). The HTML renderer runs with JavaScript off and refuses every network request except embedded data. - Identity: the worker's managed identity can reach only that container and queue.
- Container image: by default the worker pulls
realizer.azurecr.io/accesspoint/document-worker:<version>anonymously from Realizer's registry when a replica starts (one replica is kept running by default) — see Section 5.6. Customers may mirror it into their own registry (documentWorkerImage+documentWorkerRegistryResourceId) or pin a digest (documentWorkerImageDigest). The optional accessibility checker (deployAccessibilityCheck) works the same way with its own image.
4.6.8 Application Insights and Log Analytics (always deployed)
- Data processed: operational telemetry from the customer's API — HTTP method, path, status code, and duration of each request with a correlation ID; dependency calls to SQL, Blob Storage, Graph, and the optional services (duration and success/failure); unhandled exceptions with stack traces; document-conversion and media-processing timings; and the render-engine diagnostics used to confirm that email rendering is working. Telemetry may contain record identifiers and user identifiers needed to trace an operation; AccessPoint does not write document content, prompts, or requestor contact details into telemetry.
- Azure resource logs (on by default,
deployDiagnosticSettings): the same workspace also receives the Azure platform's own logs for the deployment's resources — for the API and portal web apps, a record of every request (time, client IP address, method, URL path, status, duration, browser user agent), the applications' console output, sign-ins to the publishing endpoints and access-restriction checks; for Azure Front Door, when deployed, every request at the edge with the client IP address and every web application firewall decision; and for the documents storage account, every read, write and delete with the calling identity and its IP address. For the public portal this means a requester's IP address and browser details are kept in the workspace for the retention period below, for the same security and abuse-prevention purpose as in Section 4.9. A customer can turn these logs off at deployment (deployDiagnosticSettings=false). - Location and access: the Application Insights instance and Log Analytics workspace are in the customer's subscription; Realizer has no access to them. Metric alert rules are created only when the customer supplies an alert email address, which is stored in the customer's own Azure action group.
- Retention: 90 days by default. The customer sets it at deployment with the template parameter
logRetentionDays(30 to 730 days; every deployment applies the value it is given), can change it on the workspace afterwards, and can export logs for longer keeping.
4.7 Privacy by Design
AccessPoint enforces role-based data access:
- Custodians and Contributors never see requestor personally identifiable information (PII). They work from sanitized instructions.
- Notification templates for Custodian and Contributor roles automatically strip PII tokens before dispatch.
- Detected redaction text and overlay text are served only to users who hold the requestor-PII permission; transcripts follow the recording, with speaker names masked for everyone else (Section 4.6.4).
- Machine-produced redactions (pattern matches, rule-based suggestions, AI suggestions, face detection, duplicate propagation) are stored as proposals and are excluded from every burn, export, and redline path until a human accepts them.
- Administrators control which users have access to the system and at what role level; a senior official named on a case sees that case only.
- Tenant isolation is absolute. Every query in the customer's API is filtered by tenant, the filter is unconditional, and the API refuses any state-changing call whose token was issued by a different Entra tenant. No code path reads another tenant's data.
4.8 Records Captured through Repository Connectors
The Documents workspace can also add records from systems outside Microsoft 365 — network file shares, Hyland OnBase, OpenText Content Server, an Azure Blob container, a records database, a search server — through repository connectors that the customer's administrator registers. A connector is a service the customer operates (in their own Azure subscription or on their own premises; Realizer publishes ready-to-run connectors and a software development kit, but they run in the customer's environment) or a built-in reader inside the customer's API. Every capture is initiated by a user, gated by the tenant toggle Features:ConnectorsEnabled and by the same case-level authorization as a file upload, and stored as an ordinary document in the customer's Blob Storage.
- Data flow. The user's browser never contacts a connector. The customer's API calls the connector with its own managed identity and pulls the record server-side; the record is stored as a frozen copy with its source path, the version taken, its SHA-256 hash (verified against the hash the source declared, when it declared one), and who captured it and when — the chain-of-custody record an access office needs. Whole-folder collections run as background jobs in the customer's API.
- Attribution headers. Each call to a connector carries the acting user's Entra object ID and user principal name, the case number (for example
Request:A-2026-0142), the purpose of the call (browse, search, capture, collect or test) and — only when the connector's manifest asks for them — the user's Entra group IDs. These let the source system apply its own access rules and keep its own access log; they are sent to the connector host the customer chose, and to nobody else. - What the connector sees. The connector receives the search terms and item identifiers the user acts on and the attribution above. A connector declared as application with user check may trim results per user and refuse a capture; an application-mode connector relies on AccessPoint's own gate. Requestor personal information is never sent to a connector.
- Realizer's involvement. None at run time. Realizer never sees connector traffic, search terms, captured content or attribution headers. When the customer installs Realizer's connector packages or the connector SDK from
get.realizer.io, or pulls a connector container image fromrealizer.azurecr.io, Realizer observes ordinary download metadata only (Section 5.6). - Retention. Captured records follow the retention rules of the case they belong to and are deleted with it; the provenance record is deleted with the document. A connector registration that documents were captured through cannot be deleted while those documents exist (it can be disabled), so provenance keeps resolving for as long as the record is retained.
4.9 Public Intake Portal and Online Payments (Optional)
A customer may deploy a public portal (deployPublicPortal=true, then switched on in Settings) where members of the public submit requests (and, where the customer publishes those forms, complaints or incident reports), follow their status, reply, pay fees and download responses. The portal is a separate web application in the customer's own subscription, with its own identity that can reach only the portal's own database schema and storage containers. It never contacts Realizer, and the data below is the customer's, under the customer's control.
- What a requester provides: the answers to the customer's form, their email address (the only field every form requires), and any attachments (at most 25 files, within the customer's size limits). The email address is stored encrypted with a key the portal itself cannot read, beside a one-way hash used to find the requester's submissions.
- Confirming it is them: every portal session needs an 8-digit code sent to the requester's email address (valid 10 minutes, single use, locked after 5 wrong attempts). The code email is sent by the customer's own API from the customer's own mailbox, so a copy remains in that mailbox's Sent Items. The code is kept only as a hash, except in the outgoing-mail queue, which is emptied an hour after sending.
- Cookies: the portal sets two, both functional — a session cookie (
__Host-ap-portal, only after a code is confirmed; 30 minutes of inactivity, 8 hours at most) and, only when the requester changes language, a language cookie (ap-lang, 365 days). It loads no analytics, advertising or third-party script. When the form is embedded in the customer's own web site, it runs in an isolated frame the host page cannot read. - IP addresses and browser details: recorded for security and abuse prevention with each submission, verification code, session, message and payment session, and kept for the periods the customer configures or the defaults: unconfirmed submissions about 8 days, codes and sessions 1 day after they expire, portal events 30 days, and the portal's copy of a registered submission 30 days (
Portal:RegisteredRetentionDays; 0 keeps it). Portal events are also written to the customer's tamper-evident audit ledger, which is not purged. The ledger records the email hash and, instead of the IP address, a keyed pseudonym of it — the key is replaced and destroyed every 30 days (the portal's event retention), after which no one can tell which address an entry came from — and the network the address belongs to (its first three numbers for an IPv4 address, the first 48 bits for IPv6). Ledger entries written before October 2, 2026 keep the IP address itself: the ledger's integrity check covers every entry, so they cannot be altered. With the Azure resource logs on (the default), the web server's and Azure Front Door's request logs also hold each request's IP address and browser details for the Log Analytics retention period (Section 4.6.8). - Responses: delivered through time-limited download links (30 days by default) that only a session confirmed by the same email address can open; the copies are deleted when the link expires or is revoked.
- Reading room (off by default): a customer may publish released (redacted) records. Publishing takes two different staff members; an entry never shows the request text or the requester.
- Online payments (off by default): the requester pays on the payment provider's own page — AccessPoint never receives a card number. The provider receives the amount, currency, a short reference, a description containing the request number, return addresses and language; Stripe also receives the requester's email address, a generic hosted payment page receives it only if the customer configures that, and Paymentus receives neither a name nor an email address. The provider's signed notification of the result is kept with the payment record. The payment provider is the customer's own supplier, under the customer's agreement with it.
- Edge protection (optional): Azure Front Door with a web application firewall can sit in front of the portal, in the customer's subscription.
4.10 Case Archiving: Storage Tiers, Preservation Holds and Disposition
AccessPoint can lower the storage cost of closed cases and dispose of records a records schedule no longer requires. All of it is off until the customer's administrator switches a storage policy on; the jurisdiction packs suggest one policy per case type, imported switched off.
- Storage tiers. A storage policy moves a closed case's files to a cheaper access tier — Cool, Cold, or, for original records only and only once the statutory waiting periods the customer switched on have ended, Archive — in the same storage account, region and encryption. The file is not copied anywhere else, and nothing is sent outside the customer's subscription. An archived file is restored in place when someone needs it (to view, download, reopen, hold or release it), which takes hours. AccessPoint records which files it moved and when, so it can tell users what is online.
- Housekeeping. A policy may also remove duplicate copies once a case is closed — for example, an unsent response package when a later package of the same response has been marked as sent — and derived copies (such as preview conversions) that can be regenerated from the original.
- Preservation holds. A user holding the preservation-hold permission can place a hold on a case, with a reason and an optional matter or file reference. While a hold stands, nothing on that case is deleted, purged, thinned, disposed of or moved to the Archive tier; a hold on a complaint also holds the requests linked to it. Placing and releasing a hold are recorded in the audit ledger.
- Disposition. An administrator may, case by case and never on a timer, keep only what was released (the response packages marked as sent, the originals that were themselves delivered, and the correspondence) or only the case record (the case's details, history and the hashes of what it held). Every condition is checked again at the moment of disposal: the case must be closed, past its waiting periods, free of holds and of open linked cases. The text AccessPoint extracted from a disposed file goes with it.
- The disposition log. One row per disposed item — the case number, the document name, its hash and size, and what was removed — written before anything is deleted and never purged, like the retention purge log (Section 9.1). If the case is later purged or deleted, the document names in the log are cleared; the hashes and sizes stay.
- Marking a package as sent. Staff can record that a response package reached the requester (the method, the day and an optional note); the record shows who marked it.
4.11 Discovery Surveys
The privacy program can ask staff, through a discovery survey, where personal information is held and who is responsible for it.
- What is collected: each respondent's answers — to AccessPoint's standard questions and to any questions the customer adds — any files the respondent attaches (at most 20 per answer), and the language the respondent answered in. Respondents are the customer's own staff; requesters and the public never receive a survey.
- Where it goes: the answers are stored in the customer's Azure SQL database and the files as documents in the customer's Blob Storage. The privacy office reviews the answers, and may use them to fill in or update a privacy subject's record of processing; an answer never changes a record by itself. People allowed to view the privacy program can export a survey to Excel; each export is recorded in the audit ledger.
- Translation: when AI Assist is deployed and machine translation is on, the questions the customer writes are machine-translated through the customer's own Azure OpenAI resource (Section 4.6.1) so each respondent can read them in their language; the original wording is kept beside every translation.
- Retention: a survey's files are removed 12 months after the survey closes by default — the customer can choose from 1 to 120 months, or keep them until the survey is deleted. The answers, the review decisions and the export stay and still name the files. Each removal is recorded in the audit ledger.
4.12 Lab Mode and the Training Copy
AccessPoint keeps a training copy of the customer's organization (on by default; an administrator can switch it off) where staff practice on fictional training cases in Lab mode (called train mode before version 2.6.0).
- Where it lives: in the customer's own Azure database, storage account and search index, under a separate tenant identifier (the customer's tenant ID followed by
~training), kept apart from real work by the same controls that keep tenants apart. Nothing in the training copy appears in real lists, reports, search, exports or the audit ledger, and nothing in real work appears in the training copy. - What it holds: a copy of the customer's configuration (request types, templates, letters, settings) as of its last refresh — never real cases, requesters, documents, the audit ledger, secrets, repository connectors, payment providers, portal settings or notification contacts — and the fictional training cases staff create, with invented people, organizations and addresses. Training documents are generated (the Lab library), never uploaded unless the customer allows uploads in Lab mode. The copy is refreshed on a schedule (weekly by default), which removes every training case.
- What never leaves it: letters are simulated and never sent; the requester portal, payments, repository connectors, Microsoft 365 capture, uploads and data import are refused; staff notifications about a training case reach only the people training on it, in the app and by email marked [Training], and never through Teams (so nothing reaches Realizer's Teams relay, Section 5.2).
- Records about real staff: to support training, AccessPoint records each person's training progress (an item done or dismissed, and the day — no timings, scores or clicks), learning paths assigned to them, open Lab mode sessions, team exercise participation and the debrief answers each participant writes. Progress, finished assignments and ended exercises are kept for 90 days by default (30 to 365 days, longer while an open assignment needs them); ended sessions are deleted after 7 days; a person's records are removed when they leave the organization. The worked examples in the training copy may show a decision or opinion in a real colleague's name; such records are labelled recorded by the training example.
- Removal: an administrator can remove all training data at any time (Settings → Training → Remove all training data), which deletes every row and file of the training copy and switches Lab mode off.
- Simulated people (version 2.6.0): in Lab mode's training cases, the colleagues and outsiders a trainee works with can be simulated people who answer from a script: fictional characters that exist only in the training copy, with an undeliverable placeholder address, no sign-in, no directory account and no role, never shown, listed or notified outside Lab mode. An organization's own scripts, written in AccessPoint Learning's Studio, are checked by the same rule as Realizer's, which refuses real-looking addresses and identifiers. In a team exercise a colleague may play a person instead; their answers are logged under their name like the rest of their training activity.
- No real record on a read-only licence (version 2.6.0): while the licence is read-only (Section 5.1) or Hosted AccessPoint's Free plan, uploads into the training copy are refused, whatever Settings → Training says; the Lab library's fictional records are unaffected.
- The AccessPoint Training Lab at
learn.realizer.io— Realizer's online simulation where training providers run classes — is a separate service, not part of the App: nothing passes between a customer's AccessPoint and the Lab, and when a training lead records a Training Lab certificate as training done elsewhere (Section 4.13), AccessPoint stores the link as entered and never opens it.
4.13 AccessPoint Learning (version 2.6.0)
AccessPoint Learning — learning modules the customer writes or receives in jurisdiction packs, learning paths, checks, decisions, certificates and completion records — runs entirely in the customer's own AccessPoint; nothing about anyone's learning is sent to Realizer.
- Records about staff: each person's enrolments (who assigned them, or which audience rule, the due date and status); for each step, whether it is done, when, and how AccessPoint knows (it saw it done, a named mentor confirmed it, or the person said so) — never time spent, clicks, scrolling or what was viewed; check results (latest score, whether passed, attempts); decisions and reflection notes, which are private to the person; acknowledgements; completion records and certificates (the name the person chose to print and a fingerprint of the verification code, never the code itself); more attempts a training lead granted, with the reason; and training done elsewhere that a training lead recorded (title, provider, dates, a verification link and a note).
- Who sees what: a person sees all of their own learning. Training leads see the progress and completion records of the people whose training they assigned, and only as much of check results as the customer chooses (Who sees check results: completion only, passed or not, or scores and attempts); never a decision or a note. Content statistics for authors are shown only across at least five learners, never by name.
- Retention: step states follow the training progress (90 days by default, removed when the person leaves); completion records, certificates and training done elsewhere are kept 7 years by default (1 to 15) after the later of completion and the last valid day, also after the person leaves, so a certificate can still be verified inside the organization.
- The customer's User Guide describes all of this for employee-monitoring policies (AccessPoint Learning and employee monitoring).
5. Data Processed by Realizer (Platform Services, Hosted AccessPoint and Realizer's Other Services)
For the Microsoft 365 edition, Realizer's Platform services process a limited set of data to support licensing, Teams notifications, jurisdiction packs, first-time setup, Teams tab routing, and software distribution. Email notifications, the in-app notification feed, documents, case records, AI prompts, and search indexes never transit Realizer. Hosted AccessPoint, where Realizer hosts all of that on the customer's behalf, is Section 5.9; Realizer's other services are Sections 5.10 to 5.12.
5.1 Licence Validation
When the App starts and periodically during use, the customer's API sends a request to Realizer's Platform API (api.realizer.io) to validate the customer's subscription. While a licence is invalid the API backs off (five minutes, doubling) rather than polling continuously.
Data transmitted:
- Tenant ID (a Microsoft Entra ID identifier for the customer's organization)
- API version string (e.g., "2.5.0") for compatibility checking
- The customer API's own base URL (e.g.,
https://app-accesspoint-<name>.azurewebsites.net) — self-registered with the Platform so the SharePoint web part can discover the customer's API address without manual configuration (see Section 5.3)
The request authenticates with a Microsoft Entra token issued to the customer's App Service managed identity, which carries a License.Validate role on Realizer's application; the token's tenant claim identifies the customer. Where the customer's Azure subscription belongs to a different Entra tenant from its Microsoft 365 (a split-tenant installation), the token is issued instead to the installation's Microsoft 365 application — an app registration with no secret or key, which proves itself with a user-assigned managed identity in the customer's subscription — so the token still names the customer's Microsoft 365 tenant and carries nothing more. A legacy static API key remains supported as a fallback for un-migrated deployments.
Data NOT transmitted:
- No user names, email addresses, or personal data
- No request content, documents, or customer business data
- No usage counts, case counts, or feature-usage statistics
Purpose: To verify that the customer has an active AccessPoint subscription. The response carries the signed licence, which names the modules the subscription includes (for example AI Assist, the public portal, video and audio records, repository connectors) so the App can offer what is licensed; it holds no personal data. The response also carries the latest published AccessPoint version number (and, optionally, a link to the update template) so the App can show administrators an "Update available" notice. Realizer's Platform determines that number by reading its own release marker (get.realizer.io/public/accesspoint/latest/version.txt); the customer's API does not contact the distribution site for this.
When a subscription lapses (version 2.6.0). If the subscription has expired beyond its grace period, or Realizer revoked it, AccessPoint becomes read-only instead of stopping: no data is deleted; reading, search, reports, exports and Lab mode keep working; every change is refused; and e-mail and Teams notices, the public portal's intake and payments, repository connectors and AI Assist pause until the subscription is renewed. Nothing more is sent to Realizer.
Data stored: one licence-token record per tenant and subscription (replaced on each validation; revoked records are retained as an audit trail), the tenant's subscription status, and, on the tenant record, the self-registered API base URL, the time of the first and the latest validation, and the version the latest validation reported.
How Realizer uses these records: besides licensing, Realizer uses these existing records — no new data is collected — to understand whether a deployment is set up and up to date, and to support the customer.
Retention: Tenant ID and licence status are retained for the duration of the subscription plus 90 days after expiration for billing reconciliation. The self-registered API base URL is stored as part of the tenant's subscription record.
5.2 Teams Activity Feed Notifications
Teams activity feed notifications are optional and default to on. In the default mode, when a notification is triggered (e.g., a custodian is assigned new work), the customer's API sends a request to Realizer's Platform API, which sends the notification via Microsoft Graph on behalf of the customer's tenant. This relay exists because Graph requires that activity notifications be sent by the application that owns the Teams app manifest, and AccessPoint's Teams app is published under Realizer's multi-tenant application.
Data transmitted (default relay mode):
| Field | Contains | Personal data? |
|---|---|---|
tenantId | Customer tenant GUID | No — identifier |
recipientUserId | Recipient's Entra user object ID | No — identifier |
activityType | A fixed manifest value (e.g., assignmentCreated) | No |
previewText | Assignment or task name and due date | Potentially — free text |
topicText | The in-app notification text | Potentially — free text |
templateParameters.actorName | Display name of the staff member who triggered the notification | Yes — a personal name |
templateParameters.requestNumber | Request reference number (request-anchored types only) | No — reference |
relatedEntity, relatedRecordId | Record type and GUID for the click-through deep link | No — identifiers |
Data NOT transmitted:
- No requestor PII (names, contact details, addresses) — there is no requestor field in the payload, and Custodian/Contributor notifications are PII-blanked before dispatch regardless
- No document content
- No request details beyond the notification parameters listed above
Two controls reduce or eliminate this flow:
- Minimize — the tenant setting
Notifications:TeamsMinimalPayload(Settings → Setup → Teams Notifications → "Minimize Teams notification content") replacespreviewText,topicText, andactorNamewith fixed placeholders ("AccessPoint Notification", "You have a new AccessPoint notification", "AccessPoint"). No free text or personal name leaves the tenant; only routing identifiers do. - Direct relay — the app setting
Notifications:TeamsRelayMode=Directmakes the customer's API call Graph itself with its managed identity, so nothing reaches Realizer. It requires theTeamsActivity.Sendrole on the customer's managed identity and a Teams manifest pointed at the customer's own application (User Guide, Step 8e).
Turning Teams notifications off in the Setup panel also eliminates the flow; email and in-app delivery of the same notifications are always in-tenant. Notifications about training cases (Section 4.12) are never sent through Teams.
Purpose: To deliver Teams activity feed notifications.
Data stored: none. The Platform logs the tenant identifier of each relayed notification for troubleshooting.
Retention: The Platform's operational logs, including the tenant identifier of each relayed notification, are kept for 90 days for troubleshooting, then automatically deleted.
5.3 API Address Discovery
When the SharePoint web part loads and no manual API-address override is configured in the tenant, it calls the Platform's api-discovery endpoint to look up the customer API address that the licence validation self-registered (Section 5.1).
Data transmitted: a Microsoft Entra bearer token issued to the signed-in user for Realizer's application — the same token the web part uses to call the customer's own API. The Platform reads only the token's tenant claim to select the record; it does not store the token or the user's identity.
Data returned: the customer's own API base URL.
Data stored: none beyond the tenant record already described in Section 5.1. The Platform logs the tenant ID and whether a URL was found.
5.4 Teams Personal Tab Routing and First-Time Setup
Teams tab routing. The AccessPoint Teams app's personal tab opens through a Platform address (api.realizer.io/apps/accesspoint/redirect-to-sp) that decides whether the tenant has AccessPoint installed and redirects the user's browser to their own SharePoint site (or to a welcome page when it does not). This design is required by how Teams desktop establishes single sign-on with SharePoint.
- Data transmitted: the tenant's SharePoint domain (e.g.,
contoso.sharepoint.com), the tenant ID, and the user's locale — supplied by Teams as query parameters. No user identity, token, or content is sent. - Data stored: none. The Platform logs the three values for troubleshooting and answers with an HTTP redirect.
Finish Setup page. After a template deployment, an administrator may open the Platform's Finish Setup page (api.realizer.io/apps/accesspoint/setup) to grant the Microsoft Graph permissions the App's managed identity needs (Section 8.1), instead of running the deployment script.
- Data transmitted: the tenant ID and the managed identity's object ID (from the deployment output), and the administrator's choice of whether to include Copilot capture. The administrator signs in to a dedicated multi-tenant "AccessPoint Setup" application, and the page performs the grants with the administrator's delegated Graph token — Entra ID itself enforces that the administrator is entitled to make them; the page adds no privilege the administrator does not already hold.
- Data stored: none. The request context is carried in a sealed, tamper-proof state value valid for 15 minutes; the administrator's token is used for the duration of the request and is not retained.
SharePoint package install page (2.5.0). For a new installation an administrator may open the Platform's install page (api.realizer.io/apps/accesspoint/install) to install the App's SharePoint package into the tenant app catalog instead of uploading it by hand.
- Data transmitted to Realizer: the tenant ID, the SharePoint name and the version, which are part of the page address, with the IP address and time any web request carries. The page downloads the release files it verified from Realizer; nothing about the tenant is sent with them.
- Not transmitted to Realizer: the administrator's sign-in. The administrator's browser signs in to Microsoft (a public "AccessPoint Installer" application with one delegated SharePoint permission), keeps the SharePoint token in the page's memory and sends it only to the tenant's own SharePoint. Realizer's server refuses any request that carries a token. What is installed, and by whom, is recorded by SharePoint in the tenant's own app catalog.
- Data stored: none beyond the Platform's ordinary request logs.
5.5 Jurisdiction Packs
Jurisdiction packs (jurisdiction-specific templates, exemptions, holidays, notification templates, and detection rules) are downloaded from the Platform on request by an administrator. Each download and each installation report authenticates as in Section 5.1.
- Data transmitted: the tenant ID (via the token), the pack code requested, and — on installation — the list of item types installed and the Entra object ID of the installing administrator.
- Data stored: the installation record (tenant, pack, item types, installing user ID, timestamp), used to notify customers of pack updates.
- How Realizer uses these records: besides pack updates, Realizer uses the same records — no new data is collected — to understand whether a deployment is set up, to support the customer, and to count which jurisdictions its customers work in. The installing user ID is not used for this.
5.6 Software Distribution
AccessPoint's installable artifacts are published to Realizer-operated endpoints. Fetching them reveals ordinary request metadata to Realizer — the artifact and version requested, the connecting IP address, and the time — and nothing else.
get.realizer.io— at deployment and upgrade time, the customer's Azure deployment downloads the API package for the requested version, and administrators download the deployment template, deployment script, and SharePoint package from the same site. The running App contacts this site in one case only: when an administrator chooses Update now, Reinstall a previous version or Re-submit the permission requests in Settings → Setup, the customer's API downloads that version's SharePoint package, checksum list and signatures to verify them before the administrator's browser installs the package. The request carries nothing but the file and version requested (the site observes the API's outbound IP address and the time, as for any download); no user identity, token or tenant data is sent. Otherwise there is no runtime dependency.realizer.azurecr.io— the document worker (deployed by default, Section 4.6.7) pulls its image anonymously when a replica starts, and the optional media processing worker (Section 4.6.6) and accessibility checker pull theirs each time they scale up from zero. The registry therefore observes the image tag, the egress IP address, the timestamp and — for the media worker, which runs when there is media work — a pull cadence that approximates the tenant's media-processing bursts. It never observes content. Mirroring the images into the customer's own registry removes this flow entirely.- In-app help (
realizer.io/docs) — the Help panel in the App shows Realizer's documentation site in a frame, so the user's browser requests those pages from Realizer's web site. That site observes ordinary request metadata (the page, the browser's IP address and user agent, the time); the App sends it no case data, user identity or token. - Connector packages (
get.realizer.io/…/connectors/, and connector images onrealizer.azurecr.iowhere the customer chooses a container host) — the connector SDK and Realizer's ready-to-run connectors (Section 4.8) download like any other artifact, at deployment time and by an administrator's choice. A running connector never contacts Realizer.
5.7 Summary — What Realizer Receives, and When
| Flow | When | What Realizer receives | Contains personal data? | How to eliminate |
|---|---|---|---|---|
| Licence validation | Startup and periodically | Tenant ID, API version, API base URL | No | Not applicable (required for licensing) |
| Teams activity relay | Each Teams notification (default mode) | Tenant and recipient GUIDs, activity type, request number, record ID; plus title, preview text, and acting user's name unless minimized | Acting user's name and free text unless minimized | Minimize toggle, Direct relay mode, or disable Teams notifications |
| API address discovery | Web part load (when no override and cache expired) | User's bearer token (tenant claim read; not stored) | Token only, transiently | Configure the manual AccessPoint_ApiUrl storage-entity override |
| Teams tab routing | Each time the Teams personal tab is opened | SharePoint domain, tenant ID, locale | No | Use AccessPoint from SharePoint rather than the Teams tab |
| Finish Setup page | Once per setup or upgrade, by an administrator | Tenant ID, managed identity ID, administrator's delegated token (transient) | Administrator identity, transiently | Use Deploy-AccessPoint.ps1 instead |
| SharePoint package install page | Once per new installation, by an administrator | Tenant ID, SharePoint name, version, IP address, time — never a token (it stays in the administrator's browser) | No | Install from AppSource, with Deploy-AccessPoint.ps1 -DeploySharePointApp or by hand |
| Jurisdiction packs | On administrator request | Tenant ID, pack code, item types, installing user ID | Installing administrator's object ID | Do not import packs (universal baseline is imported automatically) |
| Artifact downloads | Deployment and upgrade; and when an administrator chooses Update now (SharePoint package) in Settings → Setup | Version requested, IP address, time | No | Update the SharePoint package by hand instead of with Update now |
| Container image pull | Document worker replica start (default on); each media-worker or accessibility-checker scale-up | Image tag, IP address, time, pull cadence | No | Mirror the images into the customer's own registry |
| In-app help | Each time a user opens the Help panel | Page requested, browser IP address and user agent, time | IP address only, as for any web site | Do not open the Help panel |
| Repository connector traffic (browse, search, capture, collection; attribution headers) | Never — the customer's API talks to connector hosts the customer operates | — | — | — |
| Email notifications | Never | — | — | — |
| Account contact (Section 5.8) | Once, at subscription activation, from the Microsoft commercial marketplace | The subscription's beneficiary email address (or the purchaser's), the purchaser and beneficiary tenant IDs, and the subscription name | Yes — a work email address | Not applicable; it is how we reach you about your subscription and about security. You choose which mailbox it is |
| Hosted AccessPoint (Section 5.9) | While the organization uses it | Everything the organization keeps in AccessPoint, hosted for it as its processor | Yes | Use the Microsoft 365 edition |
| Hosted organization record and activity summary (Sections 5.9.3 and 5.9.4) | At creation; then at most once a day | Organization name, the creator's name and email address, region, sign-up attribution, the creator's Microsoft organization (or a hash of a personal account); counts and dates of use | The creator's name and email address | Not applicable (needed to run the service) |
| Hosted active-session counts (Section 5.9.4) | Every five minutes, from each region | Two counts for the whole region: sign-in sessions active in the last 5 and 15 minutes, and when they were measured | No | Not applicable (needed to run the service) |
| Hosted billing (Section 5.9.7) | Paid plans | Plan, add-ons and Stripe's subscription status; usage counts over the plan's allowance | A billing email address, held by Stripe | Stay on the trial or the Free plan |
5.8 Your Account Contact and Security Notices
What we hold, and where it comes from. When a subscription is activated through the Microsoft commercial marketplace, Microsoft's fulfilment service gives Realizer the subscription's beneficiary email address (falling back to the purchaser's), the purchaser and beneficiary tenant identifiers, and the subscription name. The address is stored on the tenant and subscription records as the organization's account contact. For Hosted AccessPoint, the account contact is the name and email address of the person who created the organization (Section 5.9.3). Apart from the members of a hosted organization, whose data Realizer hosts for the organization (Section 5.9), it is the only contact information for your organization that Realizer holds — AccessPoint's own security, privacy, and operations contacts stay inside your tenant and are never sent to Realizer (Section 5.1 stands: licence validation transmits no personal data).
What we use it for.
- Security advisories and incident notices about AccessPoint and Realizer's Platform services, including the notices Realizer commits to sending when an incident affects your deployment, your environment, or the software we distribute to you.
- Account, subscription, and legal notices, including material changes to this policy (Section 13).
- Managing your organization's account: Realizer records the organization and its subscriptions in its customer relationship management system, HubSpot — and an address there only when it is proven to be the organization's own (Section 5.12). We do not send marketing email to these addresses, and we do not sell or share them.
Why it matters that it is a role mailbox. This address is Realizer's out-of-band channel: the one that still works when your deployment is down, when it is the thing that is compromised, or when the artifacts you deploy are themselves the incident. We ask customers to make it a monitored role mailbox (for example itsecurity@yourorganization.gov) rather than an individual's address, to confirm it at activation, and to keep it current — a personal address leaves with the person, and a notice that lands in a disused inbox is the same as no notice at all.
How to update it. Contact support@realizer.io (or privacy@realizer.io) and we will change it; re-activating the subscription from the Microsoft marketplace also rewrites it. We confirm the change with the new address before using it.
What we record about the address itself. Beside the address and an optional label: whether it has been confirmed, when it was confirmed, when it was last re-confirmed, when we last sent it a confirmation or reminder, whether it looks like a role mailbox, and how the confirmation was made — either the literal marker that it was confirmed from the address itself or on the activation page, or, where marketplace sign-in is in use, the sign-in name of the person who activated the subscription. That sign-in name is the only other personal identifier in this record; it exists so that a later dispute about who set an organization's notification address has an answer.
Retention. Kept for the duration of the subscription plus 90 days, on the same schedule as the licence and subscription records (Section 9.2).
How Realizer contacts you. Security notices and advisories are sent from security@realizer.io, a shared mailbox in Realizer's own Microsoft 365 tenant, delivered through Microsoft Graph. Microsoft provides that mailbox, as it does the rest of Realizer's business systems; no separate email marketing or delivery service is involved, and no customer address is uploaded to a third-party sending platform.
- What we record about a message we send you: which organization it was sent to, the notice identifier, the subject, the channel, the time, the delivery result, and — where a notice asks you to acknowledge it — your acknowledgement. The delivery log deliberately holds no e-mail address: the one address we know is the account contact described above, and copying it onto every notice would turn an evidence log into a second store of personal data. Not the contents of any reply beyond what we need in order to answer it.
- What we will never do in a security notice: send an attachment; ask for a password, a multi-factor code, or an administrator consent grant; ask you to install anything outside your own deployment process; or ask you to sign in through a link in the message. Every notice carries a reference you can check independently — on our status page, on our advisory page, or in AccessPoint itself. If something claiming to be from us does not match this, it is not from us.
- Subscribing without giving us an address: our public advisories are available as a web page and an Atom feed, and service events — outages, degraded service and maintenance of Realizer's Platform services — are posted on our status page at
https://www.realizer.io/status(also as JSON athttps://www.realizer.io/status.json). None of them requires an account or an email address (Section 5.6 describes the request metadata a download or a page fetch reveals). The status page loads no analytics or other third-party scripts. - Location: Realizer's Microsoft 365 tenant, including the
security@realizer.iomailbox, is hosted in Canada (Microsoft's Canada Central region); message content is processed by Microsoft under Realizer's own agreement with Microsoft.
Status of this section: the account contact and the marketplace data have always been held (they were missing from earlier versions of this policy — see Section 15). The advisory page (
https://api.realizer.io/security/advisories), its Atom feed (https://api.realizer.io/security/advisories/feed.atom) and thesecurity@realizer.iosending mailbox are live (confirmed 2026-10-05). The status page (https://www.realizer.io/status) is served by Realizer's website, separately from the Platform services it reports on, so it stays reachable when they are not.
5.9 Hosted AccessPoint (app.realizer.io, version 2.6.0)
Hosted AccessPoint is the same software run by Realizer: an organization signs in with a Microsoft account and uses AccessPoint in a web browser, with no Azure subscription or SharePoint of its own. It is part of version 2.6.0. Here Realizer hosts the organization's data: everything Section 4 describes as kept in the customer's own Azure subscription is kept in Realizer's, for the organization. Realizer is the organization's service provider — a data processor — for that data; the organization remains the controller (Section 2) and decides what it collects, keeps and deletes.
The service is governed by the Terms of Service for Hosted AccessPoint, published on Realizer's website beside its general Terms; wherever they concern personal information, they follow this notice.
5.9.1 Where the data is
- One Azure subscription of Realizer's, in Canada Central: the web application that signs people in, the AccessPoint API, one Azure SQL database shared by many organizations, a separate database for sign-in sessions, Blob Storage, Key Vault, Application Insights and Log Analytics.
- Addresses: the service's Canadian region answers at
ca.app.realizer.io(and its API atapi.ca.app.realizer.io).app.realizer.iois the service's entry point, run by Realizer in Canada beside its control plane: its sign-up page shows where an organization's data will be kept before the sign-up continues at the region's own address, and signing in there finds the regions where the person belongs — it asks each region, with the person's sign-in, which organizations they are a member of — and sends them to theirs. It sees the account's sign-in details only for that request and keeps nothing: no session, no sign-in token, no list of who belongs where; its only cookie is__Host-ap-region(Section 5.9.2). Older links toapp.realizer.ioare forwarded, with their page address, toca.app.realizer.io. The data stays where this section says, whichever address a person uses. - Kept apart by organization: every row carries its organization; every query is filtered by it, and SQL Server row-level security enforces the same rule inside the database. The application connects with a least-privilege identity (Entra-only — no SQL password exists), and each organization's files sit under its own storage prefix.
- E-mail goes through Azure Communication Services in Realizer's subscription, with its data stored in Canada (Section 5.9.5).
- Azure Front Door with a web application firewall sits in front of the service. It is a global Microsoft edge service, so a request may first reach Microsoft's nearest point of presence.
- Logs and backups: web request logs (with the browser's IP address) and telemetry are kept 30 days. The databases' point-in-time restore backups, and the stored files, are geo-replicated within Canada — from Canada Central to its paired region, Canada East — so a copy survives the loss of one region and never leaves Canada. Production runs Microsoft Defender for SQL.
- Malware scanning: on, from the deployment of the service template of November 7, 2026: Microsoft Defender for Storage, in the same subscription and region, scans every file uploaded to the service. The verdict is kept with the file; a file found malicious cannot be previewed, downloaded or released, and a file still being scanned is held. Defender's sensitive-data discovery is off.
- Conversion, recordings and accessibility checks: from the same deployment, Office, web-page and e-mail files are converted to PDF, video and audio recordings are processed, and release packages are checked for accessibility by workers in Azure Container Apps in the same subscription and region. Each job handles one organization's files, the workers keep nothing between jobs, they never reach the database, and their storage permissions let them write only the files they produce — never an organization's original files.
- AI services: text recognition (Azure AI Document Intelligence), content and semantic search (Azure AI Search) and transcription of recordings (Azure AI Speech) run in the same subscription in Canada Central. AI Assist's language models (Azure OpenAI) are off for each organization until its Administrator allows them, because they may process prompts outside Canada (Section 5.9.11).
- Not part of Hosted AccessPoint: anything that needs the organization's Microsoft 365 (records captured from Microsoft 365, Teams notifications, the nightly directory check, roles bound to Entra groups) and repository connectors. The public portal is offered only with a paid plan and its add-on, and only where the service runs it; Realizer's hosted service does not run it yet.
5.9.2 Signing in and members
- Sign-in is Microsoft's. A work or school account of any organization, or a personal Microsoft account, signs in to Realizer's AccessPoint Hosted application and agrees once that it may "use AccessPoint as you". The account's own organization — or Microsoft, for a personal account — checks the password and any multi-factor authentication.
- What the service records about a member: the account's identifier (its Microsoft tenant and object ID), name, email address and preferred language, when the membership was created and when the member last signed in, the member's roles, and — like any AccessPoint user — what they do in the audit history (Section 4.2).
- Invitations. A person joins an organization only through an Administrator's invitation: the invited address, the roles offered, who invited and when. The link is kept only as a hash and works for 14 days, and only an account whose sign-in proves the invited address can accept it. An Administrator may also limit the organization to work or school accounts of listed Microsoft organizations (Who can join).
- Colleagues and requests to join. Someone signing up with a work or school account is shown the names — and only the names — of the hosted organizations of their own Microsoft organization that let colleagues find them (an Administrator can turn that off; it is on by default) and whose Who can join would admit them, and may ask to join one. A request holds what their sign-in gave: the account's identifier, an address the sign-in proves, and their name, with when it was made and who decided it. It is kept in that organization, seen by its Administrators (who also get a notice and an e-mail), approved with the usual invitation or declined; a request not decided in 30 days expires, and a decided or expired one is deleted 90 days later. Personal Microsoft accounts are never shown anything this way.
- Sessions and cookies. The sign-in tokens stay on Realizer's server, never in the browser. The browser holds functional cookies only —
__Host-ap-session(the session),__Host-ap-org(which organization is open),__Host-ap-ended(why a session ended), the sign-in flow's own, and at the entry pointapp.realizer.io__Host-ap-region(the key of the region the browser last chose or was sent to, such asca, kept a year so the next visit goes straight there; nothing about the person) — and the pages load no analytics, advertising or third-party script. The service records each session's last activity on the server so it can sign members out after the time the organization's Administrator chose (from 15 minutes to 2 hours without activity, and at most 12 hours in all, on Realizer's service); members are warned before. Each region keeps its own sessions; nothing of a session crosses a region. - IP addresses. The web application passes the browser's IP address to the API for rate limits; the web request logs keep it 30 days.
5.9.3 The organization's record at Realizer
When an organization is created, the service registers it with Realizer's control plane (api.realizer.io, part of the Platform):
- the organization's name and its region;
- the name and email address of the person who created it, kept as the organization's account contact for security and account notices (Section 5.8) and as its billing contact;
- where that person came from (sign-up attribution, Section 5.9.6);
- the organization's Microsoft organization — the tenant identifier of the work or school account that created it — or, when that was a personal Microsoft account, a one-way hash of that account's identifier (never the identifier itself): the organizations sharing it form one customer, whose free allowance and trial they share.
The shared Free allowance. For each Microsoft organization (or hashed personal account) whose organizations use the Free plan, the control plane keeps that free allowance: the day its year started and, for each request or assessment one of the organizations took from it, the kind, the number, the date and which organization took it — counts and dates only, no name and no case content. An organization learns from it only how much is left, never which other organization took what. When an organization is deleted, what it took stays counted for the rest of that year, and ninety days after the deletion it no longer says which organization took it (below); the allowance itself is deleted once none of the Microsoft organization's organizations is left and its last year has ended.
The control plane issues the organization's licence and keeps its plan, add-ons, trial and billing state and the history of changes to them (for example a trial extension, with the requesting member's masked address and reason). Jurisdiction packs are read through it, with no installation record. The service itself keeps a registry entry with the organization's name, its status, the creating account's identifier (to limit how many organizations one account creates and which of them start with the trial), the creating work account's Microsoft organization (one trial per Microsoft organization, and to show the organization's name to colleagues who sign up), whether colleagues can find it, and its Who can join list.
When the organization is deleted. When the service deletes an organization — an inactive Free organization purged, an abandoned trial removed, or an organization that leaves (Section 5.9.8) — it tells the control plane the date and which of the three it was, and nothing else. The organization's subscription ends then; a lapse or the Free plan does not end it. Ninety days after the deletion Realizer deletes its licence records of the organization, its staff's notes about it and its account contact, and removes from the records it keeps for its accounts and for the service's figures everything that names the organization or its people: the name, the creator's name and address, the sign-up attribution, its Microsoft organization (or the hash of a personal account) and the masked address of the member who asked for a trial extension — and its share of a shared Free allowance stops naming it. What stays is the plan, statuses, dates, counts and amounts under an internal identifier, and one record that the deletion happened: a one-way hash of the organization's identifier, the dates and the counts. Its company is deleted from HubSpot and its deals keep only their plan, stage and value (Section 5.12); Stripe keeps its invoices and payments for Realizer's accounts, without the customer's email address and name (Section 5.9.7). If the organization's subscription becomes active again at Realizer before then, nothing is deleted until Realizer has looked at it. The admin console's audit of what Realizer's own staff did is never edited, and can still name the organization as the subject of an action (Section 5.12).
5.9.4 The activity summary (service operations data)
With its licence refresh, at most once a day, the service reports for each organization a summary that holds counts and dates only: the dates of the first real request, the first assessment, the first Lab case finished and the second member to join; the number of members; how many signed in over the last 30 days; and the last day anyone signed in. It holds no name, no address and no case content — its format has no text field, and tests fail if one is added. Realizer keeps only the latest summary for each organization and uses it to see whether organizations get started with the service, to support them and to report on the service; it is not sent to HubSpot or to any analytics service.
Active sessions. Every five minutes, over the same connection to Realizer's control plane, each region reports how many sign-in sessions were active on it in the last 5 and 15 minutes — two counts for the whole region, with no name, address or organization — so Realizer can plan maintenance for a quiet time. A session counts as active when its member did something in the service in that time; the region counts its own sessions and sends only the two numbers. Realizer keeps these counts for eight days, to show a week of activity by hour.
5.9.5 Email
- Every email of a hosted organization — letters to requesters, notices to members, invitations, confirmation codes and Realizer's lifecycle notices — is sent through Azure Communication Services in Realizer's subscription from
accesspoint@notices.realizer.io. The recipients and the content pass through it; its data is stored in Canada. - Replies go to the address the organization's Administrator chose and confirmed with a 6-digit code sent to that address (valid 60 minutes); until it is confirmed, the organization's mail carries no Reply-To of its own.
- Budgets. The service counts each organization's recipients before a message leaves — 100 an hour and 500 a day, or 25 and 100 during the trial and on the Free plan — so the shared address cannot be used to send mail in bulk. Mail over the budget is not sent.
- Lifecycle notices are Realizer's own fixed texts, sent to the organization's Administrators in each one's language: reminders 7, 3 and 0 days before the trial ends, the trial extension, the Free plan's yearly cap, the move to the Free plan, and the purge warnings (Section 5.9.8). They come from AccessPoint, and replies go to Realizer support. Switching the organization's email notices off stops the reminders, but never the notice of the move to the Free plan or a purge warning. The service records which notice went to which organization, when, and how many people it reached.
5.9.6 Sign-up attribution
When someone reaches the sign-up page from the marketing demo or a campaign link, the page keeps only these keys from the link, in the browser's session storage, through sign-in: pack, lang, from, demo, steps, size and cta (the jurisdiction pack and language, where the visitor came from, the demo sample, the steps completed, the price band checked and the button pressed), the campaign tags utm_source, utm_medium, utm_campaign, utm_term, utm_content and utm_id, and gclid (an advertising click identifier). Each value is checked against a pattern and limited to 100 characters, in the browser, by the service and again by Realizer, which stores them on the organization's record (at most 2,000 characters) to learn which demos and campaigns bring customers. They are never used to build a link or a page, and Realizer does not send them to HubSpot, Google or anyone else.
5.9.7 Billing (Stripe)
The trial needs no card. Paid plans are billed by Stripe. Realizer creates the organization as a Stripe customer with the organization's name, its hosted identifier and a billing email address — that of the Administrator who first opens checkout. On Stripe's own checkout page the Administrator enters the card, the billing address and, optionally, a tax number; these go to Stripe and never reach AccessPoint or Realizer. Before the first purchase the Administrator chooses the organization's billing country, which sets the currency — Canadian dollars in Canada, US dollars in the United States, pounds in the United Kingdom, euros in the euro area, US dollars elsewhere — and the organization's subscription keeps it; the country and currency are kept with the organization's billing record. Stripe tells Realizer the subscription's status, plan, currency and add-ons. Requests and assessments beyond a plan's allowance are not billed by usage: an Administrator may buy a bundle of units, charged once to the card on file through a Stripe invoice. An Administrator manages billing in Stripe's own customer portal. Realizer's revenue reports keep Stripe's identifiers, statuses, dates and amounts only (Section 5.12). When Realizer deletes its records of a deleted organization (Section 5.9.3), it removes the customer's email address and name from Stripe and keeps the Stripe customer: its invoices, credit notes and payments are Realizer's accounting records. Stripe keeps them — each invoice with the name and billing address it was issued to, and the customer with the billing address and any tax number entered on Stripe's page — under Stripe's own retention, as Realizer's payment processor and for the tax and financial record-keeping that applies to Stripe and to Realizer.
5.9.8 The trial, the Free plan, a lapse and leaving
- The trial lasts 30 days. When it ends — or when a paid plan is cancelled — the organization moves to the Free plan: nothing is deleted, everything stays readable and exportable, received requests are still recorded, and the parts the Free plan does not include become read-only. An Administrator may extend the trial once, by 30 days, for procurement.
- An inactive Free organization may be deleted. If no member signs in and no case is created for 90 days after the move to the Free plan (for an organization that never really started — no real case, a single member, no Lab case finished) or for 12 months (otherwise), the organization may be purged. Every active Administrator is warned by email 30 and 7 days before — and at least 90 days before, with the number of open requests, while a received request is open — and Settings → Subscription shows the date. A warning that reached nobody is sent again and does not count. Any member's sign-in or a new case cancels the schedule, and a preservation hold prevents the purge. The purge is a deliberate action by Realizer, never automatic.
- Download all of your organization's data. An Administrator can download a ZIP of everything the organization holds — every record and file, requester details included; the Lab mode training copy is left out. Up to three a month; the link works once and for five minutes, the file is deleted after 14 days, and every start and download is recorded in the organization's audit ledger.
- Leaving. Realizer deletes an organization that leaves the service once a complete export of it has been recorded — every record and file, its training copy and its search entries — keeping one record that it existed. Deleted rows leave the database's backups as the backups age out.
5.9.9 Realizer's access
Realizer operates the service and administers its Azure resources. Operations on a whole organization — export, a move to another database, purge — run only through an operator command inside the service's own environment, never through a web page or an API, and each is logged with the operator's name. The settings that send every organization's email belong to Realizer; an organization's Administrator manages everything else. Realizer uses an organization's data only to provide, secure and support the service, and does not sell or share it.
5.9.10 Who else processes it
- Microsoft — Azure hosting in Canada Central, Entra ID sign-in, Azure Communication Services email (data stored in Canada), Azure Front Door (a global edge service), Application Insights, Azure AI Document Intelligence, Azure AI Search and Azure AI Speech (Canada Central), and — only for an organization whose Administrator allowed it — Azure OpenAI's Global deployments (Section 5.9.11), under Realizer's agreement with Microsoft.
- Stripe — billing of paid plans (Section 5.9.7).
Section 8.5 lists every service provider Realizer uses.
5.9.11 AI Assist
AI Assist (Ask AccessPoint, drafting, summaries, translation and AI suggestions such as redaction suggestions) sends the text the feature needs — for example the case's text and excerpts of its documents, including any personal information in them — to Azure OpenAI models that Realizer provisions in its subscription in Canada. Microsoft offers the current models in Canada only as Global deployments: Microsoft may process some prompts and answers in its other Azure regions worldwide.
- Off until allowed. For each organization, AI Assist that uses these models stays off until one of its Administrators reads a short, dated statement in Settings → Features & AI → AI Assist and allows it. Who allowed it, when, and the version and language of the statement are recorded in the organization's audit history and audit ledger. An Administrator can withdraw it at any time; AI that needs it stops at once. A materially changed statement pauses AI until it is allowed again.
- What stays in Canada: everything AccessPoint stores — cases, documents and the results AI produced — stays in Canada (Section 5.9.1). Microsoft does not use prompts or answers to train its models; its abuse monitoring keeps its own records under Microsoft's terms.
- Not affected: text recognition, search and transcription are processed in Canada Central and need no permission. The organization's AI use is counted against its plan's AI allowance; AI Assist logs only how much was used, never the prompts or answers.
5.10 The AccessPoint Training Lab (learn.realizer.io)
The Training Lab is Realizer's online simulation of AccessPoint, where training providers run classes. It is separate from the App: nothing passes between a customer's AccessPoint and the Lab (Section 4.12). Each provider decides what its classes collect and is the controller of its learners' data; Realizer runs the Lab as the provider's processor under the provider agreement.
- Learners join a class with its code and a display name. The Lab keeps that name; an email address only when the class asks for one, or when the learner confirms one to have a resume code sent to it (after a 6-digit code sent to that address); what the learner did in the exercises of the fictional simulation — the changes made, step states, check results, feedback, debrief answers and help requests; and when they were last active. It does not record clicks, page views or keystrokes. The join page says what is recorded, who sees it and when it is deleted.
- Codes. A learner's resume code, for continuing on another device, is never kept: only a hash of it and its last four characters as a hint. A class's join code is kept as a hash for joining and, encrypted, so that the class's instructors can show it again.
- Who sees it: the class's instructors and the provider's administrators. Realizer's staff reach a provider's classes and content only through an audited support view.
- Deletion. Learner data is deleted 30 days after the class closes — the provider may choose from 7 to 90 days — and instructors are reminded 7 days before; statistics without names are kept. A certificate of completion that a learner chose to receive keeps only what it prints (the name as confirmed, the provider, the course, the class and its dates, the exercises and their verified steps, the issue date), can be verified by anyone holding its link or QR code, and is kept until the provider deletes it — at the learner's request, for example — or leaves the Lab. When a provider leaves, its data is deleted. Database backups hold deleted rows for at most 7 more days.
- Provider members (administrators, authors and instructors) are invited by email and sign in with a one-time code sent to their address. The Lab keeps their name, address, roles, last sign-in and a log of their actions, and emails them when they sign in from a new browser.
- Course material belongs to the provider. It stays in the Lab's private storage, is never served with public caching, and never enters Realizer's source code repository, logs or telemetry.
- Safeguards. Joining is protected by a proof-of-work check and rate limits; failed attempts count against one Lab-wide budget, whose alerts carry counts only — never a code, a name or an address — and IP addresses are kept only in memory, as hashes, for those limits. The Lab sets functional
__Host-cookies only and loads no analytics. Uploads are scanned for malware. Telemetry carries keys and identifiers, never course text. - Where. Azure in Canada Central (the app, its database and private storage); email through Azure Communication Services with data stored in Canada, from
notices@mail.realizer.io. AI drafting, for a provider's authors only and only once the provider agreed to it, runs on Azure OpenAI; Microsoft's abuse monitoring may keep prompts for up to 30 days.
5.11 The Marketing Demo (demo.realizer.io)
- The demo runs AccessPoint's screens on fictional data in the visitor's browser. Nothing a visitor types, creates or redacts is saved or sent anywhere; letters are simulated and nothing can be uploaded. Closing or reloading the page discards the visit.
- Analytics on direct visits. Opened on its own, the demo counts its use with the same Google Analytics property as www.realizer.io, with the same consent choices: advertising storage is denied everywhere, and analytics storage is denied in the European Economic Area, the United Kingdom and Switzerland until the visitor chooses Allow analytics on the notice. The events are fixed keys — the sample, language and jurisdiction chosen, the steps completed and the About panel's buttons — never anything typed. Google's cookies (
_ga,_ga_…) are set fordemo.realizer.ioonly, never for every realizer.io site, so they do not reach Hosted AccessPoint or the Training Lab. The visitor's choice is kept in the browser (rz-consent), and?ga=offstops a browser being counted (kept asga-optout). Embedded in www.realizer.io's own pages, the demo loads no analytics: the website counts the visit under its own notice. - Campaign links. A link's campaign tags (
utm_*,gclid) are passed on unchanged to the links the demo opens on www.realizer.io and, through the hosted sign-up, into an organization's sign-up attribution (Section 5.9.6). - Link previews. When a demo link is shared, the demo's server writes the preview's title, description and image from the sample and language the link names — nothing about the visitor.
5.12 Realizer's Customer Records and Admin Console
Realizer's customer records (HubSpot). Realizer keeps a record of each customer organization in its customer relationship management system, HubSpot (operated by HubSpot, Inc.; Realizer's account is hosted in HubSpot's Canadian data centre in Montréal), to manage subscriptions, renewals, billing and support.
- For every customer: the organization's name and its AccessPoint tenant identifier, and a deal for each subscription — the channel (Microsoft commercial marketplace or Hosted AccessPoint), the plan, its stage (trial, paid, lapsed or churned) and its value (in US dollars for the marketplace, Canadian dollars for Hosted AccessPoint).
- A person's email address only when it is proven to be the organization's own. For a subscription bought through the Microsoft commercial marketplace: the purchaser's address, only when the purchase was made with a work or school account of the customer's own Microsoft organization — never a personal Microsoft account, and never a partner buying for the customer — and the confirmed account contact (Section 5.8) only when it is at that same organization's domain. An address at a public email or internet-provider domain is never sent, and never names or matches a company. Only the address is sent — no name or other detail.
- Hosted AccessPoint customers are recorded with no contact at all — only the organization and its deal. The person who signs up gives Realizer an address alone, and nothing proves it belongs to the organization rather than to that person. When Realizer deletes its records of a deleted hosted organization (Section 5.9.3), its company is deleted from HubSpot (HubSpot removes a deleted record for good after 90 days) when Realizer created that company for it; a company Realizer matched to an organization already in its records, or that was merged with another, is kept, and only the organization's identifier is removed from it. Either way its deals keep their plan, stage and value without its name.
- Never sent: case data, requester information, documents, anything typed in AccessPoint, the members of a hosted organization, and a hosted organization's activity summary or sign-up attribution.
- We do not send marketing email from these records, and we do not sell or share them. To ask what HubSpot holds about you, or to have it removed, write to privacy@realizer.io.
Billing and revenue records. For its revenue reports and accounting, Realizer imports Microsoft marketplace orders and revenue from Partner Center, with the customer's company name — never a person's name or email address — and, for Hosted AccessPoint, Stripe's paid invoices, credit notes and payouts, read with a read-only key, keeping Stripe's identifiers, statuses, dates and amounts — never a name, address, email or tax number. Monthly summarized journals of revenue, fees and tax are posted to Realizer's accounting system, QuickBooks Online — totals only, no personal data.
The admin console (admin.realizer.io) is Realizer's internal tool over these records. Only Realizer's own staff accounts can sign in, with multi-factor authentication; email addresses are masked in its lists; and every action is recorded in an append-only audit with who acted, what they did, why, and the record before and after. That audit is never edited: it is Realizer's evidence of what its own staff did, it holds no email address of a customer's people, and an entry about an action on an organization can still name that organization after Realizer's other records of it are deleted.
6. Data We Do NOT Collect
For the Microsoft 365 edition, Realizer does not collect, store, or have access to:
- Access request content or details, assessments, incidents, complaints, or risks
- Requestor personal information (names, addresses, contact details)
- Documents, recordings, transcripts, redactions, or response packages
- Records captured through repository connectors, connector search terms, or the attribution headers sent to a connector
- AI prompts, AI outputs, search indexes, or AI activity logs
- User browsing behaviour or device information (note: the customer's Application Insights instance collects operational telemetry within their own Azure subscription — Realizer does not have access to this data)
- Cookies or local storage identifiers
- Location data
- Usage statistics or feature-usage telemetry from the App
For Hosted AccessPoint, Realizer holds the organization's case data, documents and requester information on the organization's behalf (Section 5.9) and uses them only to provide, secure and support the service: none of it goes to Realizer's customer records (HubSpot), billing or accounting, or to any analytics, and the hosted pages load no analytics or advertising script. What Realizer receives for its own records about a hosted organization is the organization's record and a summary of counts and dates (Sections 5.9.3 and 5.9.4). The marketing demo's usage statistics (Section 5.11) are the only analytics Realizer runs around AccessPoint.
7. Data Storage and Security
7.1 Customer Data (App)
- Stored in the customer's Azure subscription (Azure SQL Database, Azure Blob Storage, and — when deployed — the optional Azure services described in Section 4.6)
- Encrypted at rest using Microsoft-managed encryption keys; customers may use their own keys instead, held in their own Azure Key Vault, for the documents storage account (a deployment parameter) and for the database (Transparent Data Encryption with a customer-managed key)
- Encrypted in transit via TLS 1.3 (App Service) and TLS 1.2 or higher (Azure SQL, Blob Storage); the storage account accepts HTTPS only and blocks public blob access
- Access controlled by the customer's Entra ID and Azure RBAC policies; database access uses Entra-only authentication via managed identity (no SQL credentials exist), Blob Storage has shared-key access disabled, and every optional service disables local authentication — no key for any AccessPoint resource exists to be leaked
- Stored secrets — connector and payment-provider credentials and the private key that decrypts requester e-mail addresses received through the public portal — are encrypted with the application's Data Protection keys, and each of those keys is itself encrypted with a key held in an Azure Key Vault in the customer's subscription (deployed by default; the vault has purge protection, and only the application's identity may use the key)
- Blob Storage keeps the previous version of every overwritten or deleted file and soft-deletes deleted files and versions for 30 days, so an accidental overwrite or deletion stays recoverable by the customer for at least 30 days. Previous versions are removed automatically 30 days after they were created, and soft-deleted copies are permanently removed 30 days after deletion (retention of deleted records: Section 9.1)
- Optional hardening at deployment: Microsoft Defender for SQL with vulnerability assessment (on by default; findings are stored in a dedicated storage account in the customer's subscription), Microsoft Defender for Storage malware scanning (Section 4.6.5), private virtual-network isolation of the database, storage and key vault (
deployNetworkIsolation) — through service endpoints, or through private endpoints with the services' public network access turned off, the AI services included (networkIsolationMode), a write-once evidence container with a time-based immutability policy (deployEvidenceWormContainer, default seven years) for exported evidence packages and audit-ledger anchors, deleted automatically a week after that period unless the customer chooses to keep them (deleteEvidenceAfterRetention), and a least-privilege database identity for the application that cannot change or delete audit-ledger or purge-log rows (sqlLeastPrivilegeRuntime) - Every write is recorded in the audit history and in a hash-chained audit ledger whose verification scheme is versioned so that historical chains remain verifiable; the chain's latest entry is recorded outside the database at least daily (in the customer's Application Insights and, when deployed, the write-once evidence container), so a rewritten chain is detected by verification
- Realizer has no standing access to customer Azure resources
7.2 Platform Data (Realizer)
- Hosted on Azure infrastructure in Canada
- Encrypted at rest and in transit
- Access restricted to authorized Realizer personnel with MFA-protected accounts
- Hosted AccessPoint stores each organization's data in Realizer's Azure subscription in Canada Central, with the safeguards of Section 7.1 that apply to a service: encryption at rest and in transit, Entra-only database access through a least-privilege identity, SQL Server row-level security between organizations, storage without shared keys, Key Vault, an audit ledger for each organization, Microsoft Defender for SQL in production, and web request logs kept 30 days (Section 5.9)
- The Training Lab stores its data in its own Azure SQL database and private storage in Canada Central (Section 5.10)
- For the Microsoft 365 edition, no customer case data is stored on Realizer infrastructure — no requests, assessments, incidents, complaints, risks, requestor information, documents, recordings, transcripts, AI prompts, or search indexes. The limited contact information Realizer does hold is the list in Section 5: the purchaser and beneficiary email addresses the Microsoft commercial marketplace supplies when a subscription is activated, the Entra object ID of an administrator who installs a jurisdiction pack, the identifiers in Teams relay transit logs (Section 5.2), and the account contact derived from the marketplace addresses (Section 5.8), which is also recorded in Realizer's customer relationship management system
7.3 Website Data
- Contact form submissions are stored securely using Microsoft Azure infrastructure hosted in Canada
- Encrypted at rest and in transit
- A tool's results email (Section 3.1) is not stored: only the one-way hash of the address is kept, in memory, for 24 hours
8. Third-Party Data Sharing
Realizer does not sell, rent, or share customer data with third parties.
The third-party services involved in data processing are:
8.1 Microsoft Graph API
AccessPoint reads and writes Microsoft 365 data through Microsoft Graph under three sets of permissions. All Graph calls are made from the customer's tenant (by the signed-in user's browser or by the customer's API) except the Teams activity relay described in Section 5.2.
Application permissions on the customer's App Service managed identity — granted by the administrator via the Finish Setup page or Deploy-AccessPoint.ps1; used by the customer's API without a signed-in user:
| Permission | Purpose | What is read or sent |
|---|---|---|
Mail.Send | Send email as the mailboxes AccessPoint is configured to use: notifications, portal verification codes and security-advisory notices from the configured shared sender mailbox; correspondence and automatic acknowledgements from the mailbox chosen for the letter — the sending officer's own mailbox (the default), an intake mailbox the customer has listed, or the shared sender mailbox | Message content, from the customer's own mailbox to the recipient through the customer's own Exchange Online — never via Realizer. Grants no read access: the API cannot open, list or search any mailbox. Tenant-wide until the customer restricts it with an Exchange application access policy to a group holding the mailboxes AccessPoint sends as |
User.ReadBasic.All | Validate that the shared mailbox exists; resolve notification recipients | Basic profile fields (name, email) only; no mailbox contents |
TeamsAppInstallation.ReadForUser.All | Check whether the AccessPoint Teams app is installed for a recipient before sending a Teams notification | The recipient's installed-apps list, filtered to AccessPoint |
Application.Read.All | Teams setup validation — confirm Realizer's application is consented in the tenant | Service principal metadata; read-only |
AppCatalog.Read.All | Teams setup validation — confirm the AccessPoint Teams app is in the organization's catalog | Catalog metadata; read-only |
AiEnterpriseInteraction.Read.All | "Add from Microsoft 365 → Copilot" capture (Section 4.5) | The signed-in user's own Copilot interaction history — the API always scopes the call to that user's object ID. Granted by default; opt out at setup (-DisableCopilotCapture, or the Finish Setup checkbox) |
User.Read.All | Directory reconciliation (nightly sweep for disabled or deleted user accounts) | Account status (accountEnabled) of AccessPoint's own users and the ids in the deleted-users recycle bin — never mail, files or message content. Granted by default; opt out at setup (-DisableDirectoryReconciliation, or the Finish Setup checkbox) |
GroupMember.Read.All | Group-based roles — resolve whether a signing-in AccessPoint user belongs to an Entra group an administrator bound to a role | Membership of AccessPoint's own users in the specific groups bound to AccessPoint roles only (checkMemberGroups, together with User.ReadBasic.All) — never the tenant's group directory, other groups' membership lists, mail, files or message content. Granted by default under the same opt-out as User.Read.All |
TeamsActivity.Send (optional) | Only when the customer adopts the Direct relay mode (Section 5.2) | Teams activity notifications sent by the customer's own identity |
The managed identity also holds the non-Graph License.Validate role on Realizer's application (Section 5.1). On a split-tenant installation (Section 5.1) these permissions and that role are held instead by the installation's Microsoft 365 application in the customer's Microsoft 365 tenant, which signs in with no secret through a user-assigned managed identity in the customer's subscription; the permissions and their purposes are the same.
Delegated permissions requested by the SharePoint package — approved by a SharePoint administrator; exercised in the signed-in user's browser with the user's own identity, so the user can never read more than they already can in Microsoft 365:
| Permission | Purpose |
|---|---|
User.Read.All | People picker, user search and resolution |
Sites.Read.All | SharePoint site browsing and Lists capture |
Files.Read.All | File browsing and capture from sites and OneDrive |
Mail.Read | Email capture and preview from the user's own mailbox |
Mail.Read.Shared | Email capture from shared and intake mailboxes the user already has access to |
Calendars.Read | Calendar capture (Section 4.5) |
Calendars.Read.Shared | Calendar capture from shared mailboxes |
Chat.Read | Teams chat capture |
ChannelMessage.Read.All | Teams channel message capture |
Team.ReadBasic.All, Channel.ReadBasic.All | Listing the user's teams and channels for capture |
Notes.Read, Notes.Read.All | OneNote capture |
OnlineMeetingRecording.Read.All | Listing and downloading recordings of meetings the user organized (Section 4.5) |
OnlineMeetingTranscript.Read.All | Downloading the Teams transcript of those meetings |
OnlineMeetings.Read | Reading the meeting metadata needed to list recordings |
GroupMember.Read.All | The group picker an administrator uses to bind an Entra group to an AccessPoint role |
access_as_user (Realizer's application) | The token audience for calls to the customer's own API and to the API address discovery endpoint (Section 5.3); not a Graph permission |
Realizer's multi-tenant application — consented by the customer's administrator; operated by Realizer for the Teams relay only:
| Permission | Purpose |
|---|---|
TeamsActivity.Send | Send Teams activity feed notifications on behalf of AccessPoint (Section 5.2). Cannot read Teams messages, channels, or any other tenant data |
AppCatalog.Read.All | Resolve the Teams app's catalog identifier so the notification carries the AccessPoint icon; read-only |
User.Read | Basic sign-in scope |
8.2 Realizer Platform API
Used to validate the subscription, relay Teams notifications, discover the API address, route the Teams tab, finish setup, serve the SharePoint package install page, and download jurisdiction packs — each described in Section 5 with the exact data transmitted — and, for Hosted AccessPoint, as the control plane that registers organizations, issues their licences and keeps their plans and billing state (Section 5.9.3).
8.3 Microsoft Azure (optional services)
Azure OpenAI, Azure AI Search, Azure AI Document Intelligence, Azure AI Speech, Microsoft Defender for Storage, Azure Container Apps, and Application Insights — when the customer chooses to deploy them, these are Azure resources inside the customer's own Azure subscription under the customer's agreement with Microsoft; document and case content processed by them does not leave the customer's Azure environment and is never sent to Realizer (see Section 4.6).
8.4 Other
For the Microsoft 365 edition, no data is transmitted to any other third-party service, with one customer-chosen exception: when a customer switches on online payments, the payment provider it contracts receives the payment details in Section 4.9. Document conversion, redaction and PDF rendering run entirely inside the customer's Azure subscription: AccessPoint's own document engine in the App Service, and the document worker Container App (queue in, blob out, no network access needed for the content) — no document content is sent externally. The document viewer is served with all of its assets from the customer's own App Service; nothing is loaded from a third-party CDN. The face-detection and region-tracking models used for video redaction ship inside the application and call no external service. The only other outbound connections the App makes are the email image fetches described in Section 4.3, the artifact and image downloads described in Section 5.6, and the user's browser loading Realizer's help pages (Section 5.6).
8.5 Realizer's Service Providers (Subprocessors)
For Hosted AccessPoint, the Training Lab, the Platform and Realizer's own records, Realizer uses these service providers:
| Provider | For | What they process | Where |
|---|---|---|---|
| Microsoft (Azure, Entra ID, Azure Communication Services, Azure Front Door, Application Insights, Azure Container Apps, Microsoft Defender for Storage, Azure AI Document Intelligence, Azure AI Search, Azure AI Speech, Azure OpenAI) | Hosting Hosted AccessPoint, its entry (app.realizer.io), the Training Lab, the Platform and the demo; Microsoft sign-in; email; Hosted AccessPoint's AI services; the Lab's AI drafting | All of Hosted AccessPoint's data and the Lab's; sign-ins; email recipients and content; the text an AI feature sends | Canada Central, with Hosted AccessPoint's backups and stored files geo-replicated to Canada East (Communication Services data stored in Canada; the Lab's AI drafting in Canada East). Hosted AccessPoint's Azure OpenAI models are Global deployments: for an organization whose Administrator allowed them, Microsoft may process prompts and answers in any of its Azure regions (Section 5.9.11). Entra ID sign-in and Front Door are global services |
| Microsoft 365 (Realizer's own tenant) | Security notices and correspondence (Section 5.8) | The account contact's address and the notices sent to it | Canada |
| Stripe | Billing of Hosted AccessPoint's paid plans (Section 5.9.7) | The organization's name, its hosted identifier and a billing email address (the address and name are removed 90 days after the organization is deleted; its invoices and payments stay with Stripe under Stripe's retention); the card, billing address and tax number entered on Stripe's page; the billing currency; bundles bought | Stripe's own jurisdictions |
| HubSpot, Inc. | Realizer's customer records (Section 5.12) | Companies and deals; a proven organization address for Microsoft marketplace customers only | HubSpot's Canadian data centre (Montréal) |
| Intuit (QuickBooks Online) | Realizer's accounting (Section 5.12) | Monthly summarized journals — totals, no personal data | Intuit's own |
| Google (Google Analytics) | The marketing demo's usage statistics on direct visits (Section 5.11) | Fixed event keys and Google's cookies on demo.realizer.io, with the request details any web request carries | Google's own |
The customer's own deployment of the Microsoft 365 edition uses none of them except Microsoft, under the customer's own agreement (Sections 4.6 and 8.3).
9. Data Retention and Deletion
9.1 Customer Data
Data retention is fully controlled by the customer:
- Administrators configure a retention period per request type, assessment type, incident type, complaint type, (for standalone risks) risk category, and privacy subject type — a privacy subject's period counts from the day it is retired. A type with no period keeps its records indefinitely — a tenant that configures nothing purges nothing.
- The built-in Retention Review lists expired records across the five case registers and retired privacy subjects, withholds any record still needed by an open case (an open linked request or complaint, or a live risk) with the reason shown, re-checks eligibility at the moment of purge, and deletes the record with every dependent row, document, converted PDF, rendition, search-index entry, and blob — whatever storage tier the files were moved to (Section 4.10).
- A case under a preservation hold is never purged, deleted, thinned, disposed of or archived until every hold on it is released (Section 4.10).
- A purge leaves only a retention purge log entry (record number, type, date) as evidence the record existed. That log is never itself purged. The disposition log of documents disposed of under a storage policy (Section 4.10) is likewise never purged; when its case is purged or deleted, the document names in it are cleared and only the hashes, sizes and dates remain.
- Discovery survey files are removed 12 months after their survey closes by default (Section 4.11); the answers stay with the survey until it is deleted.
- The training copy (Section 4.12) loses every training case at each refresh (weekly by default) or reset, and anyone can remove the training cases they started. Training progress, finished assignments and ended team exercises are kept for 90 days by default (30 to 365); a person's training records are removed when they leave the organization; Remove all training data deletes the whole training copy.
- Requestor contacts can be anonymized or purged independently of their requests.
- When a record is purged or deleted, its files are deleted together with their previous versions (Blob Storage keeps a version of every file it overwrites or deletes). Where the deployment lets the application delete versions (the default), they are deleted at the moment of the purge; otherwise the storage lifecycle policy deletes them once they are 30 days old. Deleted files and versions then remain recoverable by the customer for 30 days under Blob Storage soft delete, and are permanently removed after that — in total within about 30 days of the purge, or about 60 days where the lifecycle policy removes the versions. The retention purge log records which of the two applied. A single document deleted from a case that is kept follows the lifecycle path (within about 60 days). Files in the optional write-once evidence container cannot be deleted before their immutability window elapses. Afterwards a storage lifecycle rule deletes each of them, with its previous versions, seven days after its window has passed (Blob Storage soft delete then keeps it 30 more days), unless the customer turned that rule off at deployment (
deleteEvidenceAfterRetention=false) because its records schedule keeps evidence longer — then nothing deletes them and the customer removes them when its schedule allows. A retention purge never reaches this container. - AI activity log entries, audit history, and the audit ledger are retained until the customer removes them; they contain no document content. The audit ledger keeps the email hash of public-portal events, with a pseudonym of the requester's IP address whose key is destroyed after 30 days and the network the address belongs to — not the address itself (Section 4.9; entries written before October 2, 2026 keep the address, because a ledger entry cannot be altered) — and none of it is removed when a case is purged.
- Ask AccessPoint conversation history is private to the user who had it and can be deleted by that user at any time. A conversation about a case is deleted when the case is purged or deleted; a conversation about no case is deleted after a period without activity set in Settings → Features → AI Assist (365 days by default; 0 keeps it until its user deletes it).
- Deleting the Azure resources (SQL Database, Blob Storage, and any optional services) permanently destroys all data.
- Realizer cannot recover customer data after deletion, as it does not hold copies.
9.2 Platform Data
- Licence records: retained for the duration of the subscription plus 90 days.
- Subscription records, including the purchaser and beneficiary email addresses supplied at activation and the account contact derived from them (Section 5.8): retained on the same schedule as licence records.
- The operational logs of the Platform endpoints in Section 5, including the Teams relay (tenant identifiers and request metadata, never content): retained for 90 days in Realizer's Log Analytics workspace in Canada, then automatically deleted.
- Jurisdiction pack installation records: retained with the tenant's subscription record, on the same schedule as licence records.
- Records of security notices we sent — notice identifier, recipient address, time, and delivery result (Section 5.8): retained for 24 months, so that we can show whom we notified and when. When a subscription's records are deleted, the address in these records is masked rather than the record being removed, so the evidence that we notified your organization survives without keeping the address. Canadian privacy law requires an organization to keep a record of a breach of security safeguards for 24 months, and customers subject to Québec's incident-register rules may need to ask us about an incident for up to five years.
- AccessPoint's own security, privacy, and operations contacts are not Platform data: they are stored in the customer's own tenant, like any other configuration, and are covered by Section 4.4 and the customer's own retention decisions.
- For the Microsoft 365 edition, no customer case data is retained on the Platform.
- Hosted AccessPoint (Section 5.9): an organization's data is kept while the organization exists, under its own retention settings (Section 9.1). A lapsed organization is kept, read-only. An inactive Free organization may be purged after the warnings of Section 5.9.8, and an organization that leaves is deleted once a complete export of it has been recorded. A download of all an organization's data is deleted after 14 days. Web request logs and telemetry: 30 days. Deleted rows leave the database's point-in-time restore backups as they age out. The control plane's record of the organization (its name, the creator's name and address, its plan, billing state and sign-up attribution) follows the subscription records above: a hosted organization's subscription ends when the service deletes the organization — a lapse or the Free plan does not end it — and 90 days later Realizer deletes or anonymizes its records of the organization as Section 5.9.3 describes, keeping one record of the deletion (a one-way hash of the organization's identifier, the dates and the counts). The activity summary is replaced by each new one; after the deletion it keeps only its counts and dates under an internal identifier. The regions' active-session counts (Section 5.9.4) are kept eight days.
- The Training Lab (Section 5.10): learner data 30 days after a class closes (7 to 90 days, the provider's choice); certificates of completion until the provider deletes them or leaves the Lab; a provider's data is deleted when it leaves; backups hold deleted rows 7 more days.
- The marketing demo (Section 5.11): the demo itself keeps nothing; its Google Analytics data is kept under the website's Google Analytics settings.
- Realizer's customer records (Section 5.12): HubSpot records follow the subscription records above (for a hosted organization: the company Realizer created for it deleted — or, for one it matched to an existing company, the organization's identifier removed — and its deals without its name, 90 days after the organization is deleted); imported billing and marketplace figures are kept as Realizer's accounting records, and hold no person's name or address.
9.3 Website Data
Contact form submissions and demo requests are retained for the duration of the business relationship. You may request deletion at any time. Nothing is kept from a tool's results email except that hash, for 24 hours (Section 3.1).
10. Data Subject Rights
10.1 For End Users of the App
Requests related to personal data processed within AccessPoint (access, correction, deletion, portability) should be directed to the customer organization that deployed AccessPoint, as they are the data controller.
10.2 For Customer Organizations
As the data controller, customers can:
- Export every register — including requestor contacts, case activity, redaction schedules, and a complete inventory of stored documents with their storage paths and hashes — from Settings → Export, and copy the underlying files directly from their own storage account.
- Export or delete any data in their Azure SQL database and Blob Storage at any time.
- Use the Retention Review feature to manage data lifecycle.
- Decommission the entire environment by deleting their Azure resources.
10.3 For Website Visitors and Platform Data
You have the right to:
- Request access to your personal information
- Request correction or deletion of your data
- Withdraw consent for communications at any time
- Lodge a complaint with a data protection authority
To exercise these rights, or to request access to, correction of, or deletion of the limited data Realizer processes (tenant ID, licence records, subscription and account contact addresses, pack installation records), contact privacy@realizer.io.
If your work address is the account contact for a subscription (Section 5.8) and you would rather it were not, tell us and we will replace it with the address your organization nominates — we will ask the organization for a replacement first, because a subscription with no reachable contact cannot be told about a security incident.
10.4 For Hosted AccessPoint, the Training Lab and the Demo
- Members of a hosted organization, and requesters and others whose information it holds, should ask that organization, which is the controller; Realizer helps it answer. Its Administrators can download all of its data at any time (Section 5.9.8).
- Learners in the Training Lab should ask the training provider that ran the class; the provider can delete a class's learner data, or a certificate, at once.
- Visitors to the demo can decline analytics on the notice, use
?ga=off, or use Google's own opt-out. - Anyone whose address Realizer holds in HubSpot or as an account contact can write to privacy@realizer.io.
11. International Data Transfers
- Customer data remains in the Azure region chosen by the customer during deployment. No cross-border transfer occurs unless the customer configures geo-redundant storage. Two optional services deserve attention: Azure OpenAI's default Global standard deployment type keeps data at rest in the account's region but allows Microsoft to run inference in other Microsoft regions (choose Data zone standard to bound it to the EU or US data zone), and from version 2.6.0 a new Azure OpenAI account is placed in a region of the same geography where the customer's region does not offer Realizer's recommended models (Section 4.6.1); and Azure AI Document Intelligence may be provisioned in the nearest supported region when the customer's region does not offer it (Section 4.6.3). The media processing worker, Speech, Search, Defender, and telemetry resources are provisioned in the customer's own region.
- Platform data (tenant ID, notification metadata, routing and setup metadata) is processed on Realizer's Azure infrastructure in Canada. The account records Realizer keeps in its customer relationship management system (Section 5.8) are held by HubSpot, Inc. in its Canadian data centre (Montréal). Software artifacts and the container image are served from Realizer-operated distribution endpoints. Customers in the EU or other jurisdictions should ensure this is compatible with their data transfer requirements; the Direct relay mode (Section 5.2) and mirrored container images (Sections 4.6.6 and 4.6.7) remove the two recurring runtime flows.
- Hosted AccessPoint stores every organization's data in Canada (Azure Canada Central; Communication Services data in Canada). Microsoft's sign-in and Azure Front Door are global services, and Stripe processes billing in its own jurisdictions (Section 8.5).
- The Training Lab stores its data in Canada Central.
- The marketing demo's analytics are processed by Google in its own jurisdictions.
- Website data is processed on Azure infrastructure in Canada.
12. Children's Privacy
AccessPoint is an enterprise business application. It is not directed at children and does not knowingly collect personal data from children under the age of 16. The same holds for the Training Lab, which is for professional training, and for the demo.
13. Changes to This Policy
We may update this Privacy Policy from time to time. Material changes will be communicated through:
- An updated "Effective Date" at the top of this document
- A notice on the Realizer website
- Notification to active customers via email
14. Contact
For questions about this Privacy Policy or AccessPoint's data practices:
Realizer Services Inc. Email: privacy@realizer.io Website: realizer.io/privacy
15. Change History
- November 14, 2026 — AccessPoint 2.6.0 and Hosted AccessPoint are released, so the revisions of November 3 to 10, 2026 below are in effect. Nothing else changes: no new data is collected.
- November 12, 2026 (this website) — Added emailing yourself a tool's results (Section 3.1) and the button and page names carried by links to the sign-up and deployment pages (Section 3.2).
- November 10, 2026 (this website) — Brought this page into line with the product policy's revisions of November 3 to 10, 2026 (the entries that follow), which add Hosted AccessPoint (Section 5.9), the AccessPoint Training Lab (Section 5.10), the marketing demo (Section 5.11), Realizer's customer records and admin console (Section 5.12), AccessPoint Learning (Section 4.13) and Realizer's service providers (Section 8.5), with what they mean for storage, retention, rights and transfers (Sections 7.2, 9.2, 10.4 and 11); those parts take effect when version 2.6.0 is released. Section 3.2 now also says how the demo is counted: opened directly at demo.realizer.io, it uses the same Google Analytics property and consent choices as this site and counts only anonymous use of its screens, never anything typed in it; inside this site's pages, this site counts it.
- November 10, 2026 (version 2.6.0, effective when it is released) — Hosted AccessPoint: the Free plan's allowance belongs to the Microsoft organization (or personal Microsoft account) and its organizations share it; Realizer's control plane keeps that shared allowance — its year and, for each request or assessment taken, the kind, the number, the date and which organization took it, counts and dates only — an organization learns only how much is left, a deleted organization's share stops naming it ninety days after the deletion, and the allowance is deleted once the Microsoft organization has no organization left and its last year has ended (Section 5.9.3).
- November 9, 2026, second revision (version 2.6.0, effective when it is released) — Hosted AccessPoint:
app.realizer.iois now the service's entry, where people sign up and sign in before being sent to their region, with the__Host-ap-regioncookie (Sections 5.9.1, 5.9.2); AI Assist runs on Azure OpenAI's Global deployments, off for each organization until its Administrator allows it, with text recognition, search and transcription in Canada Central (Sections 5.9.1, 5.9.10, 5.9.11); billing in the currency of the billing country and bundles instead of usage billing (Section 5.9.7); the regions' active-session counts for planning maintenance — two counts per region every five minutes, kept eight days (Sections 5, 5.9.4 and 9.2); the service providers' table (Section 8.5). - November 9, 2026 (version 2.6.0, effective when it is released) — Hosted AccessPoint's Canadian region has its own address,
ca.app.realizer.io;app.realizer.ioforwards every visit there and stores nothing (Section 5.9.1). Where the data is kept does not change. - November 7, 2026 (version 2.6.0, effective when it is released) — Added Hosted AccessPoint (
app.realizer.io), where Realizer hosts an organization's data in Canada as its processor: where the data is, signing in and members, the organization's record at Realizer, the activity summary of counts and dates (service operations data), email through Azure Communication Services with budgets and a confirmed Reply-To, sign-up attribution, Stripe billing, the trial, the Free plan, the purge of inactive Free organizations, Download all of your organization's data, leaving, Realizer's access and its subprocessors (Sections 1, 2, 4, 5.7, 5.9, 6, 7.2, 8.2, 8.5, 9.2, 10.4 and 11), and how long Realizer keeps its own records of a hosted organization: 90 days after the service deletes the organization, Realizer deletes or anonymizes them, keeping a one-way hash of its identifier with the dates and counts; its HubSpot company is deleted, and Stripe keeps its invoices without the customer's email address and name (Sections 5.9.3, 5.9.7, 5.12, 8.5 and 9.2). Rewrote what Realizer's customer records in HubSpot hold: a person's address only when it is proven to be the organization's own, and none for Hosted AccessPoint customers (Sections 5.8 and 5.12). Added the AccessPoint Training Lab (Section 5.10), the marketing demo and its analytics (Section 5.11), Realizer's billing and accounting records and its admin console (Section 5.12), and a list of Realizer's service providers (Section 8.5). A lapsed subscription now makes AccessPoint read-only instead of stopping it (Section 5.1). Lab mode's simulated people, and no uploads into the training copy while the licence is read-only (Section 4.12). Hosted AccessPoint's uploads are scanned for malware, and its conversions, recordings and accessibility checks run in Azure Container Apps in the same subscription and region, from the deployment that adds them (Section 5.9.1). The Microsoft 365 edition sends Realizer nothing new. Updated where the Microsoft 365 edition's AI account may be placed (Sections 4.6.1 and 11) and what Section 5.10 and 11 say about the Training Lab's AI drafting (November 8, 2026). - November 3, 2026 (version 2.6.0, effective when it is released) — Used 2.6.0's names: Lab mode (was train mode), the Lab library and learning paths (Section 4.12); described the AccessPoint Training Lab at
learn.realizer.ioas a separate service the App never contacts (Section 4.12); and added AccessPoint Learning — what it records about staff, who sees it and how long it is kept (new Section 4.13). - October 23, 2026 — Corrected the
Mail.Sendrow in Section 8.1, which named only notification emails from the shared mailbox. The permission also sends portal verification codes and security-advisory notices from that mailbox, and correspondence and automatic acknowledgements from the mailbox chosen for the letter: the sending officer's own mailbox, a listed intake mailbox, or the shared sender mailbox. It grants no read access, and a customer can limit it with an Exchange application access policy to the mailboxes AccessPoint sends as. No new data is collected, and nothing is sent to Realizer. - October 5, 2026 (status page) — Service events are posted on Realizer's public status page at
https://www.realizer.io/status, served by Realizer's website separately from the Platform; like the advisory page and feed, it needs no account or email address and loads no third-party scripts (Section 5.8). - October 5, 2026 (version 2.5.0) — Described the new data of version 2.5.0: case archiving — storage tiers within the customer's own storage account, housekeeping, preservation holds, disposition and the disposition log that is never purged (new Section 4.10; Sections 4.2, 4.3 and 9.1); discovery surveys — staff answers, attached files, the answer language, machine translation of the customer's questions, and the removal of a survey's files 12 months after it closes by default (new Section 4.11; Sections 4.6.1 and 9.1); and train mode — the training copy of the organization with fictional cases under a separate tenant identifier, what never leaves it, the progress records kept about staff and how long (new Section 4.12; Sections 4.6.1, 5.2 and 9.1). Added assessment validation records (opinions, decisions, section evaluations, regulator consultations), privacy audit contacts and privacy program records to the case data (Section 4.2); privacy subject retention to the retention rules (Section 9.1); the licensed modules carried by the signed licence (Section 5.1); the keyless identity of a split-tenant installation, which validates the licence and holds the Microsoft Graph permissions in place of the managed identity (Sections 5.1 and 8.1); the delegated
GroupMember.Read.Allpermission the SharePoint package requests for the role group picker (Section 8.1); the SharePoint package install page among the Platform's services (Sections 5.4, 5.7 and 8.2 — an optional page used once per new installation, which receives the tenant ID, SharePoint name and version but never a token); and customer-managed keys for the database (Section 7.1). Apart from that page, version 2.5.0 sends no new data to Realizer. - October 2, 2026 (SharePoint package updates) — The API now downloads a SharePoint package from
get.realizer.iowhile running, but only when an administrator chooses Update now, Reinstall a previous version or Re-submit the permission requests in Settings → Setup, to verify it before the administrator's browser installs it; the request carries only the file and version, and nothing else leaves the tenant (Sections 5.6 and 5.7). The check for Microsoft 365 permissions waiting for approval runs in the user's browser and sends nothing anywhere. - October 2, 2026 — Corrected against the product as built (version 2.4.0): transcripts now follow the recording, with speaker names masked for users without the requestor-PII permission (Sections 4.6.4 and 4.7); face detection is on by default wherever recordings are used and runs in the media worker when it is deployed (Section 4.6.6); Office conversion is done by the document worker, which is deployed by default and pulls its image from Realizer's registry (new Section 4.6.7, Sections 5.6 and 5.7); the Teams relay logs the tenant identifier only, and the Platform's operational logs are kept for 90 days, not the 30 previously stated (Sections 5.2 and 9.2); the evidence container holds evidence packages and audit-ledger anchors (not redaction masters) and nothing deletes them after their window (Sections 7.1 and 9.1); the email paragraph now says plainly that a sender's tracking pixel registers the image fetch at conversion (Section 4.3). Added the public portal and online payments (new Section 4.9: what a requester provides, the email code, the two cookies, IP-address retention, deliveries, the reading room, and what each payment provider receives), the in-app help pages loaded from Realizer's web site (Section 5.6), Realizer's customer relationship management system (HubSpot, hosted in Canada) for account management (Sections 5.8, 7.2 and 11), and stated that Realizer's own Microsoft 365 tenant, from which security notices are sent, is hosted in Canada Central (Section 5.8), and what a purge keeps (Section 9.1). Later the same day, with the deployment template changes of that date: evidence-container files are now deleted automatically a week after their immutability window unless the customer keeps them (Sections 7.1 and 9.1 — superseding "nothing deletes them after their window" above); the Log Analytics workspace now also receives Azure resource logs by default, including web-request and Front Door logs with client IP addresses and storage access logs, and its retention is a deployment parameter (
logRetentionDays, 30 to 730 days) rather than a module default (Section 4.6.8); and network isolation can use private endpoints with public network access turned off (Section 7.1). Also later that day, after a product change: the audit ledger no longer records a portal requester's IP address — new entries carry a keyed pseudonym and the network instead, while entries already written keep it (Sections 4.9 and 9.1) — and Ask AccessPoint conversations are now deleted with the case they are about, and after a configurable period without activity (365 days by default) when they are about no case (Sections 4.6.1 and 9.1). - September 27, 2026 — Stated that Realizer also uses records it already holds — licence validations (Section 5.1) and jurisdiction pack installation records (Section 5.5) — to understand whether a deployment is set up and up to date, to support the customer, and to count which jurisdictions customers work in. No new data is collected: the App sends nothing it did not send before, the Teams tab routing and Finish Setup flows still store nothing (Section 5.4), and no usage or per-user information is recorded. Section 5.1 now also names what the tenant record already kept from licence validation: the time of the first and the latest validation, and the version the latest one reported.
- September 15, 2026 (second revision) — Documented the account contact Realizer has always held but never described: the subscription's beneficiary (or purchaser) email address supplied by the Microsoft commercial marketplace at activation, what it is used for, why it should be a monitored role mailbox, how to change it, and how long it is kept (new Section 5.8, with a row in the Section 5.7 summary and the retention list in Section 9.2). Described the Realizer shared mailbox that sends security notices, what a genuine notice will never ask for, and the advisory feed and status page that can be followed without giving Realizer an address. Confirmed the boundary in Section 5.1: AccessPoint's security, privacy, and operations contacts stay inside the customer's tenant and are not sent to Realizer; licence validation still transmits no personal data. Corrected two further statements against the code as built: the notice-delivery log holds no e-mail address (Section 5.8 previously said it recorded the address), and Section 5.8 now describes the verification metadata kept beside the address, including the sign-in name of the person who activated the subscription where marketplace sign-in is in use. Corrected Section 7.2, which said no personal information at all is stored on Realizer infrastructure: no customer case data is, but the marketplace purchaser and beneficiary addresses, pack-installation administrator IDs, and Teams relay transit identifiers described in Section 5 always were, and Section 9.2 now says so.
- September 15, 2026 — Corrected the retention of deleted files (Sections 7.1 and 9.1): with Blob Storage versioning on, a deleted file was kept as a previous version that soft delete does not remove; purges now delete previous versions (or the storage lifecycle policy removes them 30 days after creation), so purged content is permanently removed within about 30 days, or about 60 days where the lifecycle policy applies. Described the Key Vault that encrypts the application's Data Protection keys, the external anchoring of the audit ledger, and the optional least-privilege database identity (Section 7.1).
- September 10, 2026 — Added Section 4.8 on records captured through repository connectors (connector hosts operated by the customer, the attribution headers each call carries — acting user's object ID and UPN, case number, purpose, group IDs on request — server-side capture with SHA-256 provenance, and Realizer's absence from the flow); listed connector packages under software distribution; added the connector row to the summary table and to the list of data Realizer does not collect.
- September 3, 2026 — Added a complete inventory of Microsoft Graph permissions (application, delegated, and Realizer's application) with the purpose of each; described every Microsoft 365 capture source and the token it uses, including the new Teams meeting recording and transcript capture; expanded the optional Azure services section to cover Azure OpenAI (including the premium deployment and deployment-type residency note), Azure AI Search, Document Intelligence OCR, Azure AI Speech transcription, Microsoft Defender for Storage, the Container Apps media worker and its container image, and Application Insights / Log Analytics, each with identity model, retention, logging, and opt-out; clarified what the AI activity log records and which AI outputs are stored as case data; documented the Teams relay field list and the two controls that minimize or eliminate it; added the API address discovery, Teams tab routing, Finish Setup, jurisdiction pack, and software distribution flows with a summary table of what Realizer receives; described in-tenant face detection and region tracking; updated the retention section for the five-register retention engine and the purge log; and added residency notes for the optional services.
- July 22, 2026 — Previous revision.