Canada's Privacy Commissioners Set Out What Clinics Owe Patients When They Use AI

news · healthcare · privacy

Canada's Privacy Commissioners Set Out What Clinics Owe Patients When They Use AI

On October 29, 2026 Canada's privacy regulators published a resolution on AI in health care: what clinics, hospitals and AI vendors are now asked to do.

Remolda Team6 min read

On October 29, 2026, Canada's federal, provincial and territorial privacy commissioners and ombuds published a joint resolution on the responsible use of artificial intelligence in health care and health research. Adopted in Ottawa on September 15, 2026, it lists what governments, AI vendors and the clinics and hospitals that buy AI systems are expected to do, from patient notice and opt-out to vendor contracts, shadow AI and human review of outputs.

For a clinic that already runs an AI scribe, a booking assistant or an inbox triage tool, the resolution reads as a checklist that privacy regulators across the country now share. Remolda's AI consulting for healthcare in Canada starts with a one-week AI Readiness Review at $490 CAD plus applicable taxes: a score across six dimensions and a written list of what to fix first. Send us your question or call +1 (613) 867‑7767.

What the privacy regulators published on October 29, 2026

The Office of the Privacy Commissioner of Canada (OPC) published the resolution, issued jointly by the federal, provincial and territorial privacy commissioners and ombuds. The resolution was passed at their annual meeting, hosted in Ottawa by the Privacy Commissioner of Canada and the Information Commissioner of Canada.

The regulators ask that AI systems in health meet six core principles through their whole life cycle, from design and procurement to use, monitoring and decommissioning: privacy protective, valid and reliable, safe, human rights affirming, transparent, and accountable. The resolution gives chatbots and AI scribes as examples of generative AI tools in health.

They make four requests of federal, provincial and territorial governments:

  1. Review, amend or adopt legislation and regulations so that AI systems in health meet the six principles.
  2. Consult the privacy regulator in their jurisdiction when they draft those frameworks.
  3. Engage with the public, including patients and people from marginalized communities.
  4. Legally require all relevant parties who develop, procure and use AI systems in health to establish AI governance and accountability frameworks based on robust risk assessments.

The resolution is a call to action addressed to governments, vendors and providers. A clinic's legal duties today come from the privacy laws that apply where it operates, such as PHIPA in Ontario (our healthcare AI privacy checklist lists them by province); the fourth request shows the direction regulators want governments to take.

"[…] governments, regulators, and service providers must work together to ensure that these technologies are developed, deployed, and used with strong privacy and data protection frameworks to protect the personal health data of Canadians." Philippe Dufresne, Privacy Commissioner of Canada

Ten expectations for clinics and hospitals that use AI

The resolution addresses health care providers that procure, adopt and use AI systems. The left column follows its wording; the right column is how we would turn each point into a document or a control in a clinic.

What the resolution asks of providersWhat it looks like in a clinic
Policies on collection, use, disclosure and retention comply with the privacy law of the jurisdictionAn AI section in the privacy policy, checked against your provincial act
Clear communication to patients where consent is required, and where practicable an opt-out before deploymentA short patient notice for the scribe and a way to decline it
A designated contact person and an accessible way to challenge consequential decisions or inferencesA named privacy lead and a written route for complaints about AI use
Limit personal information to what the purpose needsRecording limited to the visit, retention period set in the tool
Consider privacy, security and AI-specific assessments of prospective vendorsA vendor file reviewed before you sign
An AI risk management framework, monitored through the life cycle, built into broader risk management and able to inform PIAs and threat risk assessmentsOne register of AI tools and their risks, reviewed on a fixed schedule
Contracts that cover privacy and security training, record keeping, auditing and limits on use, including a ban on unauthorized commercialization of informationContract clauses checked against this list, subcontractors included
Plain explanations of how information is handled and how individuals exercise their rightsA patient-facing page or handout
Staff training and measures against AI used without approval or oversight ("shadow AI")An approved-tools list and a rule on personal AI accounts
Human oversight and verification of outputs, including human-in-the-loop reviewA clinician reviews and signs every AI-drafted note before it enters the record
Clinic AI accountability cycleFive steps a clinic repeats for each AI tool: assess the vendor, sign a contract with privacy terms, inform patients and offer an opt-out, have a clinician review outputs, then monitor risks and return to the vendor assessment.1. Assess the vendorPrivacy, security and AI-specific reviews2. Contract with privacy termsAuditing, records, limits on use3. Inform patientsNotice, opt-out, named contact person4. Clinician reviews outputsHuman-in-the-loop before the record5. Monitor risksRisk register, shadow AI checks, repeat
The provider expectations in the October 2026 resolution, arranged as a cycle a clinic repeats for each AI tool.

What to ask your AI scribe or chatbot vendor

The resolution also sets expectations for AI developers and service providers in health. They convert into six questions for the next vendor meeting:

  1. How was the system trained, and how does it work? Vendors are asked to provide evidence of training and a general explanation of how the system uses information.
  2. Can you show that our use complies with our privacy law? The resolution asks vendors to demonstrate this to the client.
  3. Will you share your risk assessments? Vendors should share ongoing findings and tell clients which mitigations to apply on their side.
  4. Do you use de-identified or synthetic data where possible, and do you know where training data came from? Both are part of the privacy-by-design list, along with data minimization and lawful sourcing.
  5. Are the most protective privacy settings on by default, and can the system be audited and tested in real conditions? The resolution lists both.
  6. Can our patients challenge an output? Vendors are asked to support human-in-the-loop oversight and a way for clients and end users to challenge AI outputs.

A first month for a clinic already using AI

A practical order for a clinic with one to three AI tools in use. The timing assumes vendors answer within a few days; a larger organization or a slow vendor will stretch it.

  1. Week 1: inventory. List every AI tool staff use, including personal accounts and browser extensions. This is where shadow AI shows up.
  2. Week 2: vendor files. Send the six questions above and compare each contract with the contract row of the table.
  3. Week 3: patients. Publish a short notice, name a contact person and decide how a patient declines the scribe.
  4. Week 4: control. Write the human review rule, open the risk register and set the next review date.

Our AI compliance in Canada service maps each AI use to the law that applies to it and produces the documents above. For clinic-specific tools and provincial rules, see our guide to AI for Canadian clinics. To start, book a call or write to us.

Official references

  • Office of the Privacy Commissioner of Canada: FPT privacy commissioners and ombuds call for the responsible use of AI in health and health research (October 29, 2026)
  • Resolution: Responsible use of artificial intelligence in health and health research (adopted in Ottawa, September 15, 2026)
Share

Related insights

Frequently asked questions

Talk to an AI transformation consultant

A 30-minute consultation (€100): you describe the situation, we tell you what to do first and what it would cost.

30 minutes, €100. English or French.