It turns out 60% of approved medical AI systems that automatically summarize consultations mixed up drug names, and some recorded symptoms that did not exist

An Ontario audit tested 20 approved AI medical scribe tools using simulated patient visits and found serious accuracy problems.

12 tools mixed up medication names, 9 added treatments or symptoms that were never spoken, and 17 missed key mental health information.

The findings highlight major risks to patient safety, clinical decision-making, and privacy when AI-generated medical notes are not tightly reviewed.

The report also raises concerns about how these tools are approved, monitored, and integrated into healthcare workflows.