Most of what clinics hear about “AI compliance” is vague enough to be useless. Manitoba’s actually isn’t. The College of Physicians & Surgeons of Manitoba published real, specific guidance—Advice to the Profession on the Responsible Use of Artificial Intelligence in the Practice of Medicine—and it comes down to four things.
Explicit patient consent. Not implied, not assumed because a patient didn’t object. A patient needs to actually know an AI tool is listening and agree to it before it starts.
Mandatory physician review. An AI-generated note is never the sole source of truth for a patient’s record. Someone with a license has to actually read it, correct it, and sign off—every time.
Indigenous data sovereignty. CPSM specifically flags OCAP principles—ownership, control, access, and possession—as a live consideration. If your clinic serves First Nations patients or communities under a data-governance agreement, that agreement matters here, not just whatever a vendor’s privacy policy says.
Facility permission. Deploying an AI tool in a practice isn’t just an individual physician’s call if your facility has its own approval process.
None of this is a checklist a vendor can tick for you. It’s a description of what responsible use actually looks like—and it’s the reason Tasrif won’t start recording until a patient’s consent is on record. More on exactly how that works in post five.