An auditor for the Ontario, Canada government found that AI agents tasked with turning doctor/patient conversations into structured notes routinely hallucinated false treatments, replaced drug names with entirely different drugs, and missed crucial information
