UK Watchdog Warns of Patient Safety Risks from AI Medical Transcription Errors
Healthwatch England has warned that AI medical transcription tools deployed in the NHS are making critical errors, such as misdiagnosing conditions and mixing up medications. These errors, often caused by AI hallucinations or accents, are frequently being caught by patients rather than the medical professionals themselves. While the UK government promotes AI scribes to reduce administrative burdens on NHS staff, uncorrected transcription errors could permanently contaminate patient records and lead to incorrect treatments. Furthermore, the UK regulator's decision not to classify these tools as medical devices leaves a critical gap in safety oversight. Currently, 27 different AI medical transcription tools are in use across England, but doctors warn that checking AI-generated notes for errors actually negates any time-saving benefits. Research shows these tools are particularly prone to errors when handling complex medical histories, multiple speakers, or patients whose first language is not English.
## BACKGROUND
AI medical scribes, also known as ambient clinical intelligence (ACI), use voice recognition and natural language processing to automatically document doctor-patient conversations. While studies show they can save clinicians thousands of hours of documentation time and reduce burnout, they rely on generative AI which is susceptible to "hallucinations"—generating plausible-sounding but completely fabricated information.