Ontario’s Auditor General found 60% of approved AI Scribe systems inserted incorrect drug information into patient notes across 20 vendors evaluated.
Key Takeaways
12 of 20 AI Scribe systems mixed up prescribed drugs; 9 of 20 fabricated treatment suggestions never discussed in recordings.
17 of 20 systems missed key mental health details; 6 missed mental health issues fully or partially.
Procurement scoring was badly skewed: domestic Ontario presence weighted at 30%, while medical note accuracy counted for only 4%.
Bias controls, privacy/risk assessments, and SOC 2 compliance each contributed 2-4% to total evaluation score.
OntarioMD recommends manual physician review, but no approved system includes a mandatory attestation feature.
Hacker News Comment Review
Commenters question the framing: without a baseline human error rate for drug transcription, the 60% figure lacks comparative context.
Anecdotal reports from enterprise LLM note-taker deployments echo the audit findings, with hallucinated commitments and missed nuance causing real organizational friction.
