Systèmes d'IA cliniques au Canada : déployés avec des lacunes documentées en matière de preuves et des violations de la vie privée

AI systems are in clinical use in Canadian healthcare — for virtual care, stroke detection, clinical documentation, and decision support. Provincial privacy investigations and a national health technology assessment have issued findings on these systems.

Alberta's Information and Privacy Commissioner issued two investigation reports (P2021-IR-02 and H2021-IR-01) on TELUS Health's Babylon virtual care platform, with 31 findings and 20 recommendations. The platform was promoted by provincial health services as a virtual care option. The investigations found that the platform used facial recognition for identity verification without notification or consent meeting the requirements of Alberta's Personal Information Protection Act and Health Information Act, shared personal health information with third-party service providers in the United States and Ireland without disclosing this to patients, and retained audio and video recordings of patient consultations beyond what the Commissioner determined was necessary for the stated purposes. TELUS Health Babylon launched the service before the OIPC had completed its review of the mandatory privacy impact assessments that had been submitted.

In September 2024, an Otter.ai AI notetaker bot autonomously joined a virtual hepatology rounds meeting at an Ontario hospital, recorded physicians discussing seven patients by name — including diagnoses and treatments — and emailed the transcript to 65 people, including a former physician who had left the hospital in June 2023. The bot joined via this physician's personal Otter.ai account, which was linked to a personal email calendar that still contained the recurring meeting invite. Ontario's Information and Privacy Commissioner investigated (HR24-00691). The hospital blocked AI scribe tools on its network. In January 2026, the IPC issued sector-wide guidance on AI scribes in healthcare.

CDA-AMC (formerly CADTH), Canada's national health technology assessment body, assessed RapidAI — a Class III medical device licensed by Health Canada for stroke detection. The assessment found no evidence meeting its review criteria on effects on patient harms, mortality, health-related quality of life, length of hospital stay, or cost-effectiveness. The expert review panel (HTERP) recommended that sites already using RapidAI continue to do so alongside clinician interpretation of imaging, but stated it could not recommend for or against new implementation at sites not already using the system. RapidAI is licensed for clinical use in Canadian hospitals.

Health Canada's regulatory framework for software as a medical device exempts software that is "only intended to support" clinical decision-making and is "not intended to replace clinical judgment." The Canadian Medical Protective Association has stated that physicians have "limited guidance on evaluating or mitigating the risks associated with AI tools" and that "a comprehensive regulatory framework for AI remains a work in progress."