NHS watchdog warns AI medical transcription tools pose patient safety risks

Healthwatch England has issued a warning regarding the use of artificial intelligence (AI) transcription tools within the National Health Service (NHS), noting that these systems may pose significant risks to patient safety. The watchdog highlighted instances where AI incorrectly recorded diagnoses and medication names, potentially leaving permanent errors in medical files.

Specific cases include an AI system falsely recording a patient as having demyelination—a condition linked to multiple sclerosis—when her scans actually showed no such issue. In other instances, the technology confused similar-sounding drug names or omitted critical instructions, such as requests for prescription refills. These errors were frequently identified by patients rather than medical professionals.

Concerns have also been raised regarding regulatory oversight, as the Medicines and Healthcare products Regulatory Agency (MHRA) has decided not to classify these AI scribes as medical devices. This decision means the technology may lack comprehensive, nationwide supervision. While the technology aims to reduce administrative burdens, some clinicians warn it may act as a ‘double-edged sword’ by requiring doctors to spend extra time auditing AI-generated notes for inaccuracies.