Clinical notes and AI in Quebec: what the professional orders allow

Writing notes takes up a large share of a clinician’s day, and tools that can listen to a consultation and turn it into a structured note have multiplied. The question is no longer whether AI can help, but what the professional orders allow, and on what conditions.

We went to the source and read what the Collège des médecins du Québec (CMQ), the Ordre des infirmières et infirmiers du Québec (OIIQ) and the Quebec government have published. Here is what they say, and what they do not. This article is neither legal nor ethics advice: when in doubt, your order remains the reference.

The CMQ collects its recommendations on AI scribes in a document on its website, listed by date. The most recent, dated January 14, 2026, covers two points it considers essential: certification and explicit consent.

Certification. To ensure the security and confidentiality of personal information, the Collège asks every physician who uses an AI scribe to choose a tool certified by Santé Québec’s certification office. If the application is not certified and the vendor does not intend to seek certification, the physician will have to reconsider the choice, according to the Collège.

Privacy impact assessment. Santé Québec has carried out such an assessment, valid for the use of AI scribes in its establishments. The CMQ specifies that any physician using an AI scribe outside an establishment, for example in a GMF, must make sure an assessment has been done for that workplace, since the technology environment may differ.

Explicit consent. The Collège reminds physicians that the patient’s explicit consent is mandatory before an AI scribe is used. It includes a discussion of the tool’s purposes, benefits, risks and limits. Implied consent, for example from a poster in the waiting room, is deemed insufficient.

In a clarification dated June 18, 2025, the CMQ states that at first use the physician must obtain free and informed consent, in particular regarding the recording of the conversation and the risks to confidentiality. The physician must explain how the patient’s information will be used by the company selling the tool: recording, electronic transmission, training of the software, server location and so on.

At later visits, a reminder is enough, and it gives the patient a chance to withdraw consent. The Collège recommends obtaining fresh consent in two situations: when a more sensitive topic will be discussed, and when a major technology change increases the risks. Its own example speaks for itself: if the patient’s data are now used outside Canada to train the scribe, the patient must consent again.

On April 2, 2025, the CMQ reminded physicians that the quality of notes always remains their responsibility: they must review the notes produced by the scribe and make the corrections needed for them to meet the Collège’s standards.

The Quebec government says the same on its page about the AI transcription program: the professional remains fully responsible for the content of the patient’s record and must always validate generated text. The page also states that the health ministry considers consent can be, at a minimum, verbal, while noting that professional orders may recommend explicit written consent. For a physician, it is the Collège’s position that sets the bar.

With funding from Canada Health Infoway, Quebec is offering about 2,000 licences for AI transcription solutions in primary care, free of charge for up to 12 consecutive months. According to the government page, the licences expire no later than December 31, 2026, and each product must obtain certification under the global verification kit. A privacy impact assessment must also be carried out by the health organization acquiring the solution.

The relevant statute is the Act respecting health and social services information, whose section 106 governs that assessment, as the CMQ points out. We cover the practical side in our privacy impact assessment guide.

In March 2025 the OIIQ released an opinion from its strategic watch committee on the impact of artificial intelligence on public protection and the nursing profession. It is built around four directions: recognize AI’s impact on practice while taking responsibilities into account, develop nurses’ skills, better regulate the use of AI with partners, and support public knowledge. Its president, Luc Mathieu, summed up the spirit: in health care, AI must remain a decision-support tool.

As of this writing, we found no OIIQ rule specific to nursing notes drafted with AI. The general obligations on documenting nursing care continue to apply.

As for the Ordre des psychologues du Québec, we found no published position dealing specifically with AI. The regulation on psychologists’ record keeping remains the framework for any note, whether written by hand or with a tool. For other professions the prudent rule is the same: check with your order before deploying a tool.

For physicians’ notes, the path is clear: a scribe certified by Santé Québec, a privacy impact assessment for the workplace, documented explicit consent and systematic review. An in-house tool, however well hosted, does not replace that certification.

Sovereign AI has its place elsewhere in the clinic: correspondence, scheduling, searching internal protocols, preparing administrative documents. For those uses, hosting the model on your premises or in your region, under a known jurisdiction, simplifies the assessment the law requires and keeps health information out of reach of foreign law. Our healthcare page describes this approach, and our article on the CLOUD Act explains why server location alone is not enough.