Legal
Privacy Policy
Flow Health is used by clinics to document patient care and prepare medical billing claims. This policy explains what information we handle on a clinic's behalf, where it is stored and processed, who it is shared with, and how you can ask questions about it.
- Effective
- 2026-10-21
- Last updated
- 2026-10-21
01Who is responsible for your information
Flow Health AI INC. operates Flow Health, software that clinics and practitioners use to document patient encounters and prepare and submit medical billing claims.
The clinic or practitioner using Flow Health is the custodian of the patient information in their workspace. They decide what is collected, why, who may see it, and how long it is kept. We are a service provider: we handle that information on the custodian's instructions and to deliver the service. For our own purposes, we keep aggregated, de-identified operational statistics — figures that cannot identify a patient, practitioner, or clinic — to operate and improve the service; see “Why we handle it” below. We also keep records of each account's use of AI features — feature, time, outcome, and volume, with no clinical content — to manage capacity and cost, and we maintain a directory of BC physicians drawn from publicly available sources. We do not sell personal information, and we do not use patient information to market anything.
If you are a patient, the clinic that treated you is your first point of contact for questions about your record, including requests to see or correct it. This policy explains what we do with that information on the clinic's behalf.
We handle personal information in accordance with Canada's Personal Information Protection and Electronic Documents Act (PIPEDA) and, for clinics in British Columbia, the Personal Information Protection Act (BC PIPA). Clinics also carry their own professional and provincial health-record obligations, which flow down to us through the agreement between us.
02Information we handle
What we hold depends on which parts of the service a clinic uses.
Patient and clinical information
- Patient identity and demographics: name, Personal Health Number or similar identifier, date of birth, and the sex or gender information a claim requires.
- Encounter information: the date and time of a visit, the clinical sections a practitioner records, and the notes written or drafted in them.
- Billing information: fee items, diagnostic codes, service dates, claim notes, submission results, and remittance outcomes.
- Documents and images uploaded to support a claim or an encounter, including patient face sheets processed to extract fields.
- Encounter audio recorded through the in-app dictation and scribe features, the transcripts produced from it, and the draft documentation and coding suggestions generated from those transcripts.
- Contact details, medical record number, the practitioner's private notes and status for the patient, MSP coverage check results, private-pay bills, vital signs, and, for hospital rounding, admission date, alternate-level-of-care status, and handover summaries.
- Photos, PDFs, Word or text files, and screen captures a practitioner attaches to an encounter as context, and the AI-generated summaries of them.
Account and workspace information
- Practitioner account details: name, email address, and the practitioner or billing profile a user maintains.
- Workspace membership and role, and the access decisions that follow from them.
- Sign-in and session information, including authentication tokens issued to keep a user signed in.
- Interface settings and preferences.
- MSP practitioner and payee numbers, Teleplan credentials, an optional profile photo, authenticator-app enrolment, and records of AI feature use.
Operational information
- Audit records of significant actions taken in the service (see audit records).
- Technical logs used to run and troubleshoot the service. Our application's own technical logs are restricted by an allowlist to values such as counts, statuses, and internal record identifiers. Our hosting providers' request logs also record IP addresses, browser type, and the addresses requested.
Information about other health professionals
- Names, College registration numbers, MSP practitioner numbers, licence status, and practice type of BC physicians, and specialist referral details from Pathways, used to look up referring and receiving practitioners.
03Why we handle it
We handle information to:
- create, review, and manage clinical documentation and billing claims;
- transcribe encounter audio and produce draft documentation and coding suggestions for a practitioner to review;
- extract information from uploaded documents so it does not have to be retyped;
- submit claims to, and retrieve remittance results from, the provincial billing system;
- operate workspaces, accounts, roles, and access control, and provide notifications and service status;
- keep the audit records that let a custodian review who accessed what;
- secure the service, investigate incidents, and provide support;
- keep records of each account's use of AI features — feature, time, outcome, and volume, with no clinical content — to manage capacity and cost;
- maintain a directory of BC physicians drawn from publicly available sources, used to look up referring and receiving practitioners;
- produce aggregated, de-identified operational statistics — usage volumes, error rates, and processing times that cannot identify a patient, practitioner, or clinic — to operate and improve the service; and
- meet legal, regulatory, and record-keeping obligations.
We collect only what the service needs for these purposes. Patient names, health numbers, dates of birth, and patient search terms are never placed in web addresses; addresses contain only random internal record identifiers and dates.
04Where information is stored and processed
Canadian data residency is a design requirement of this product, and most of it is enforced in the software itself rather than by policy alone. The following are configured to Canadian regions. The service refuses to start if its AI or transcription settings name a non-Canadian region, and its readiness checks fail if a storage location is outside Canada:
- the backend application that handles every read and write of clinical information, and the background jobs it runs — Montréal;
- stored documents, uploaded images, encounter audio, and billing artifacts — Montréal or Toronto;
- AI processing of clinical content and speech-to-text transcription of encounter audio — Montréal.
Some components are not covered by that guarantee, and we would rather say so plainly than claim more than we can support:
- The database and authentication service. The managed database that holds the clinical record, together with the authentication service, runs in Canada (Montréal), AWS ca-central-1.
- Web application delivery and sign-in. The hosting provider that serves the web application and processes the sign-in callback is not pinned to a Canadian region, and may run outside Canada. It handles session cookies, sign-in codes, the user's practitioner profile and workspace memberships, and page addresses that contain internal record identifiers and dates. Clinical content — names, health numbers, notes, audio, and documents — does not pass through it: the browser sends and receives clinical content directly to and from the Canadian backend.
- Stored billing credentials. Teleplan credentials are held in Google Cloud's secret store, which Google replicates without a fixed region.
Account and session information is personal information even though it is not a clinical record, so this is a genuine cross-border transfer. Information stored outside Canada may be accessible to foreign courts and law-enforcement agencies under the laws of that country.
05Service providers
We use a small number of service providers to run Flow Health. Each is bound to handle information only as needed to provide its service.
| Provider | What it handles | Where |
|---|---|---|
| Google Cloud | Backend application hosting, stored documents, encounter audio, billing artifacts, background job processing, and credential storage. | Canada (Montréal; storage Montréal or Toronto), except credential storage, which Google replicates without a fixed region |
| Google Cloud Gemini Enterprise Agent Platform (formerly Vertex AI) | Drafting of clinical notes, visit summaries, discharge instructions and rounding handovers; billing and diagnostic code suggestions and claim review; form suggestions and pre-fill; summaries of earlier records; and reading uploaded documents, photos and screen captures. | Canada (Montréal) |
| Google Cloud Speech-to-Text | Transcription of encounter audio after a recording is saved. | Canada (Montréal) |
| Supabase | The database holding the clinical record, user authentication and sessions, and account emails (sign-in confirmation, password reset, and security notices when a password or authenticator changes). | Canada (Montréal), AWS ca-central-1 |
| Vercel | Delivery of the web application and processing of the sign-in callback. Session cookies, sign-in codes, the user's practitioner profile and workspace memberships, and page addresses containing internal record identifiers and dates; no clinical content. | Not pinned to Canada |
| Expo | Delivery of mobile app updates and app builds. Each update check sends the device platform, app version, update channel, and a random installation identifier; no account or clinical information. | United States |
Claims are submitted to, and remittance results retrieved from, Teleplan, the British Columbia Ministry of Health's billing system. Teleplan is the payer's system rather than our service provider: information sent to it is handled under the Ministry's own rules.
We do not use analytics, advertising, crash-reporting, or product-telemetry services, and no tracking code runs in the web application. Two components of the mobile apps contact their providers: Expo's update service (see the table above), and, on Android, Google's ML Kit library, which sends Google device model, operating-system and app version, and performance data — never images or text.
06Transcription and AI-assisted documentation
Flow Health uses automated processing to: transcribe encounter audio; draft clinical notes, visit summaries, discharge instructions and rounding handover summaries; summarize a patient's earlier records for a new note; suggest billing and diagnostic codes, review claims, and suggest and pre-fill forms; read uploaded documents, photos and screen captures to extract or summarize their content; and turn a sample note into a reusable template.
A template's example note is stored with the template and sent to the AI service each time a note is drafted from it, so an example must use invented details, never a real patient's note.
Everything these features produce is a draft for a person to review. Generated note text is written into the encounter as a draft for the practitioner to review and correct, and documents read by the service can create patient and encounter records and draft claims for the practitioner to check. Suggested codes become part of a claim, and suggested form entries are kept, only when the practitioner adds or saves them. A claim is submitted only when a person submits it, and no clinical or payment decision is made automatically.
Transcription and AI processing run in Google Cloud in a Canadian region, using the configuration described above. Content is sent to these services in the form it is needed for the task, which means clinical content and patient identifiers may be included; it is not de-identified beforehand.
Under Google Cloud's terms, Google may not use the content we send to Vertex AI or Speech-to-Text to train or improve its models, and processes it only to provide the service to us. We do not use patient or clinical information to train or fine-tune any AI model. Google may hold content sent to Vertex AI in memory for up to 24 hours to speed up processing. If Google's automated abuse checks flag a request, Google may keep it for up to 90 days, and authorized Google staff may review it to check for misuse of the service. In both cases the content stays in the same Canadian region.
On the mobile apps, reading a patient label or face sheet with the camera happens entirely on the device, using Apple's Vision framework on iPhone and Google's ML Kit library built into the Android app. The photo is not sent to us or anyone else; only the details the practitioner accepts are saved, and vital signs read from a face sheet are added to the encounter's notes.
While a recording is in progress, short audio fragments are sent for transcription so the transcript keeps pace with the visit. Those fragments are deleted from storage once they have been transcribed; any left over are removed by a daily clean-up once they are a day old, so within two days. Saved recordings are kept with the encounter and deleted automatically 1 year after they are made. They are kept so the practitioner can check the note against the conversation. A practitioner can delete a recording sooner, for example if a patient withdraws consent, or choose an automatic-deletion period from 30 days to 2 years.
Patient consent
Under BC's Personal Information Protection Act, a clinic needs a patient's express consent before using an AI scribe. Our Terms of Service require every practitioner who uses Flow Health to:
- ask the patient, or their authorized representative if the patient cannot decide, out loud or in writing, before recording and before adding information from other records, such as text pasted from an electronic medical record. One conversation can cover both. Documents used only to fill in billing or registration details, such as a scanned face sheet, do not need this consent. A minor who is capable of deciding consents for themselves. A poster or notice on its own is not consent;
- first explain what the AI scribe does, each purpose for which their information will be collected and used, that Flow Health processes it in Canada as the clinic's service provider, how long recordings are kept, and who to contact with questions;
- ask anyone else present whose voice may be recorded, such as a family member or interpreter;
- note the patient's decision in their chart; and
- ask again before using a new feature that changes how patient information is used.
Patients can say no, or change their mind at any time — even partway through a visit — and their care will not change. The practitioner will stop recording and document the visit another way, and can delete the recording if the patient asks. The recorder reminds the practitioner to ask for consent; obtaining and documenting it is the clinic's responsibility, and Flow Health does not record it.
Patients with questions or concerns about the AI scribe should contact their clinic first. They can also complain to the Office of the Information and Privacy Commissioner for British Columbia.
07How we protect information
Access control
- All clinical information is read and written through our backend, which checks identity, workspace membership, and role on every request. Browser code never queries clinical tables directly.
- As a second safeguard, the database refuses browser sessions any direct access to the tables that hold clinical information, so clinical information can be reached only through the backend and its checks.
- Workspaces are separated structurally: every record carries its workspace, and records are linked by compound keys so a reference across two workspaces cannot be formed.
- Each workspace belongs to one practitioner. Patient identity details (name, health number, date of birth) are shared within the workspace; everything else — encounters, notes, documents, billing, and saved contact details — is visible only to the practitioner who owns it. Administrative roles do not grant access to another practitioner's records.
Accounts and sessions
- Accounts are created for each practitioner by us; there is no public sign-up.
- Our production authentication configuration requires passwords of at least twelve characters that combine upper and lower case letters, digits, and symbols.
- The web application signs a user out after thirty minutes of inactivity, with a warning two minutes beforehand.
- Sensitive actions, such as changing billing credentials, require the user to have authenticated within the previous fifteen minutes.
- A user can revoke all of their own active sessions, though a session already active on another device may continue for up to an hour, and our production authentication configuration rotates refresh tokens on use.
Transmission and storage
- All traffic is encrypted with TLS, and both the application and the backend instruct browsers to use HTTPS only.
- Responses carrying clinical information are marked not to be cached, so they are not written to browser or intermediate caches.
- Stored information is encrypted at rest by the managed cloud services described above.
- Credentials are held in a managed secret store, never in application code.
- A Content Security Policy with a per-request token restricts what the application is permitted to load or connect to.
- Live update notifications carry only record identifiers and event types — never names, identifiers such as a Personal Health Number, clinical text, or record contents.
No system can promise absolute security. These are the safeguards we maintain; they are reviewed as the service changes.
08Audit records
The service keeps an audit ledger so a custodian can review who did what. Audit records are append-only: the database rejects any attempt to edit or delete one, and it also refuses to delete a workspace or user account that audit records refer to.
Each record captures the acting user, their workspace and role, the action, the record it concerned, and when it happened. Audit records are written for, among other things:
- opening a patient, an encounter, or a claim, and looking a patient up by Personal Health Number;
- creating, updating, or archiving patients, encounters, and claims;
- uploading a document, issuing a document download, and downloading it;
- playing back encounter audio, viewing a transcript, and generating AI-assisted documentation;
- submitting claims and checking patient eligibility, and requesting or downloading exports;
- denied access attempts, session revocation, and changes to workspace access settings;
- every change to a clinical or billing record; printing a note or handover; rounding-list changes; changes to billing credentials and remittance requests; publishing or withdrawing a community template; automatic deletion of recordings; and retention actions.
To be accurate about the limits: audit coverage is deliberately record-level rather than exhaustive. Opening a record is logged, and each distinct list or search request is logged once, recording the page and the number of results but never the search terms. The same user repeating the same open or the same request within five minutes is not logged again. Audit records do not include IP addresses or browser details.
Audit records hold only an allowlisted set of values, such as counts, statuses, and internal record identifiers. Names, Personal Health Numbers, dates of birth, contact details, diagnoses, and transcript text are prohibited from them, and the service raises an error rather than writing a value outside that allowlist. Because record identifiers can be linked back to a person, we treat audit records as confidential and restrict access to them.
09Retention and disposal
The clinic decides how long its records are kept. Clinical records are subject to professional and provincial record-retention requirements that apply to the custodian, and we keep information for as long as the clinic requires it and our agreement with them remains in force.
We do not automatically delete clinical records. Encounter audio is deleted automatically only as set out below. The service maintains per-record disposal controls — a disposal date, a legal-hold flag, and an audited disposal process — and those controls currently run in reporting mode: they report what would be due for disposal without destroying anything. Disposal is carried out on documented instruction, and every disposal writes an audit record.
- Archiving a patient marks the record inactive rather than destroying it immediately. Removing a recording deletes its audio from storage within a day, unless the recording is under a legal hold; its transcript and notes are kept.
- Short audio fragments used to keep transcription current during a recording are deleted from storage once transcribed; any left over are removed by a daily clean-up once they are a day old, so within two days.
- Saved recordings are deleted automatically once they are 1 year old. A practitioner can choose a period from 30 days to 2 years, or remove a recording sooner. A recording under a legal hold, placed at the practitioner's request, for example because of a complaint or legal proceeding, is kept until the hold is lifted. Transcripts and notes are kept.
- Audit records are kept separately from the records they describe and are not deleted.
On termination, records are returned or disposed of as set out in the Terms of Service.
11Your rights and how to exercise them
Subject to applicable law, you may ask to see the personal information held about you, ask for a correction, ask how it has been used and disclosed, and complain about how it has been handled.
If you are a patient, direct your request to the clinic that treated you. The clinic is the custodian of your record and holds the relationship with you; we assist the clinic in responding but cannot release a clinic's records to a third party on our own initiative.
If you hold a Flow Health account, you may contact us directly about your own account information using the details in contact us. We respond to access requests within thirty days, or tell you why we need more time.
If you are not satisfied with our response, you may complain to the Office of the Privacy Commissioner of Canada, or — for a clinic in British Columbia — to the Office of the Information and Privacy Commissioner for British Columbia.
13Children
Flow Health is professional software. Accounts are created by us for practitioners and are not available to children.
Where a clinic provides care to a child, that child's record is held in the clinic's workspace like any other patient record, and the clinic remains the custodian. We do not collect information directly from patients of any age; information reaches us through the clinic.
14Privacy incidents
If we become aware of a privacy or security incident affecting information in our care, we contain it, investigate what happened, assess whether it creates a real risk of significant harm, and record the incident and our response.
We notify the affected clinic without unreasonable delay so the custodian can meet its own obligations to patients and regulators, and we report to regulators ourselves where the law requires it.
15Changes to this policy
We update this policy when the service, our service providers, or the law change. The date at the top of the page shows when the text was last revised.
For a change that materially affects how information is handled, we notify clinics before it takes effect, so a custodian can assess it against their own obligations.
16Contact us
Our Privacy Officer is responsible for our compliance with privacy law. Questions, access requests, and privacy complaints can be sent to the Privacy Officer at info@flowhealthai.com.
