An Ontario government auditor flagged that an AI transcription tool deployed for use by physicians hallucinated content and introduced factual errors into clinical records, according to a CBC News report cited in the audit findings. The tool and vendor were not identified in the available sourcing.
The audit represents a documented government finding — not an isolated complaint — placing the failure in a formal accountability record. Errors in clinical transcription carry direct patient safety implications, as inaccurate records can affect diagnosis, medication decisions, and care continuity.
Auditor-documented failures of this type create concrete reference points for regulators and procurement bodies. Health ministries and hospital networks evaluating similar tools now have a named jurisdictional precedent to cite when drafting performance requirements, accuracy thresholds, or liability clauses into vendor contracts.
Builders deploying AI in clinical workflows should treat this case as a signal that hallucination rates acceptable in general-purpose applications will face a different standard when auditors, not users, are doing the measuring. Operators running ambient documentation or transcription products in regulated healthcare environments should audit output accuracy against source recordings before those reviews happen externally.