Ontario Doctors Defend AI Scribes Used by 5,000 Physicians as Audit Flags Accuracy Risks
Updated
Updated · Toronto Star · May 23
Ontario Doctors Defend AI Scribes Used by 5,000 Physicians as Audit Flags Accuracy Risks
6 articles · Updated · Toronto Star · May 23
Ontario doctors said AI-generated visit notes are always reviewed, corrected and signed by a physician before entering a patient record, pushing back on an auditor general report that cited hallucinations and missing information.
Shelley Spence’s audit said Ontario rolled out AI scribes without proper evaluation and identified accuracy, privacy and security gaps, including wrong medication notes, omitted mental-health details and unnecessary blood-work orders.
About 5,000 Ontario doctors use the tools, and researchers say they save roughly three hours a week in after-hours administrative work while helping physicians stay more focused during appointments.
Doctors using the systems said errors have declined, poor audio and text prediction still cause some mistakes, and physicians can switch products if a scribe produces too many hallucinations.
Health Minister Sylvia Jones’s office said there have been no known reports of patient harm in Ontario, while doctors urged patients to ask how the tools work and confirm notes are reviewed.
Ontario's AI approval valued location over accuracy. How will the province now guarantee medical AI is actually safe for patients?
Your doctor's AI is always listening. Who truly owns your private medical conversations, and how can you be sure they remain confidential?
Ontario’s AI Scribe Audit Uncovers Systemic Failures: Every Approved Vendor Produced Dangerous Errors
Overview
A recent audit by Ontario's auditor general revealed serious flaws in AI medical scribes used in healthcare. The audit tested 20 government-approved vendors with simulated patient-doctor conversations and found that every vendor had issues with accuracy or completeness. Many AI scribes generated inaccurate, incomplete, or even fabricated information, such as hallucinating patient details or incorrectly transcribing data. These errors could lead to harmful treatment plans and put patient safety at risk. The findings highlight the urgent need for better oversight and stricter standards to ensure that AI tools in healthcare are safe and reliable.