Your doctor’s AI notetaker may be making things up, Ontario audit finds
Summary
Ontario's auditor general found AI medical scribes used in government programs frequently generate incorrect, incomplete, or hallucinated notes that could affect patient care. Tests across 20 approved vendors revealed issues including hallucinated referrals, incorrect prescriptions, and missed mental health details, with an average accuracy score of 12/20. The evaluation showed accuracy accounted for only 4% of a vendor's overall score, prompting calls for stronger testing and mandatory clinician review of AI-generated notes before they are logged.