AI Regulation Tracker  /  Professional guidance

Ontario's CPSO Tells Physicians: Get Patient Consent Before AI Records the Visit

The College of Physicians and Surgeons of Ontario says doctors must obtain patient consent before an AI tool records a clinical conversation, and must personally check every AI-generated note or recommendation for accuracy. The physician, not the software, stays accountable.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

Ontario's medical regulator has put its expectations for artificial intelligence in writing, and the message to doctors is direct: consent comes first, and accountability never leaves the physician. The College of Physicians and Surgeons of Ontario (CPSO) published Advice to the Profession titled "Using Artificial Intelligence in Clinical Practice," last updated in August 2025 and current for 2026. It arrives as ambient AI scribes, tools that listen to a visit and draft the clinical note, move quickly into everyday practice.

Consent before the recorder starts

The advice is explicit that physicians must obtain patient consent before recording conversations using AI. The College frames this as an extension of duties doctors already carry. Under the Personal Health Information Protection Act (PHIPA), physicians are required to obtain consent before collecting, using, or disclosing a patient's personal health information, and an AI tool that captures a clinical conversation is doing exactly that.

The practical reading is that consent should be secured and documented before any AI recording begins, and revisited as the technology or the way a practice uses it changes. Reporting on the guidance has emphasized that this consent should be express and captured rather than assumed, and that the appropriate approach can vary with the type of encounter. The College's own text stresses the consent requirement without spelling out a single script for every visit, so physicians should treat a clear, recorded consent as the safe floor and adapt it to the clinical context.

The physician stays accountable

The second load-bearing point is accountability. The advice states that physicians are ultimately accountable for their use of AI tools, including when they use AI to support clinical decision-making or medical documentation. That accountability is operationalized through a review duty: physicians need to review all information generated using AI for accuracy and completeness.

In plain terms, an AI scribe that drafts a note does not transfer responsibility for that note. A decision-support tool that flags a diagnosis or a drug interaction does not become the decision-maker. The doctor reads, checks, corrects, and owns the output. A convincing but wrong AI summary that goes unreviewed is the physician's error, not the software's.

What the advice does not do

This is guidance, not a new statute, and it is worth being precise about its reach. It is an Advice to the Profession document that interprets and applies existing CPSO expectations and PHIPA obligations to AI; it does not create a separate AI licensing regime, approve or ban specific products, or set fixed penalties. It does not tell physicians which vendor to use or certify any tool as compliant. What it does is set the professional standard against which a physician's conduct can be measured, which means a doctor who skips consent or rubber-stamps an unreviewed AI note is exposed to a professional-conduct problem, not just a technical one.

The cross-border read

For clinicians and health systems outside Ontario, the CPSO advice is instructive rather than binding. It does not govern a physician practicing in the United States. But it lands on the same fault lines that US clinicians already face with ambient AI: recording a patient conversation raises consent questions that, in several US states, run through two-party and wiretap-style consent rules, and the professional duty to verify AI-generated documentation is a standard any licensing board can be expected to apply. Any clinician deploying an ambient AI scribe, in Toronto or anywhere else, should assume that consent to record and human verification of the output are the two non-negotiable steps.

What to do now

Practices using or piloting AI should map their workflow to two checkpoints: a documented consent moment before recording, and a human review gate before any AI-generated note or recommendation is relied upon. Those two steps track the advice and, conveniently, track the privacy and accountability principles that regulators in multiple jurisdictions are converging on.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.