AI technology designed to transcribe patient-doctor consultations is generating incorrect drug names and diagnoses, according to a warning from Healthwatch England. The NHS watchdog stated that these errors could endanger patients if they are not detected and corrected before entering medical records. Healthwatch England, the statutory patient champion for the NHS, has received "multiple stories from patients who have noticed these errors when a health professional hasn't."
One incident highlighted by Healthwatch England involved a woman who was wrongly informed she had demyelination, a serious nerve condition linked to multiple sclerosis, in an AI-generated summary of her MRI scan. The scan actually showed "null demyelination," indicating the condition was not present. The patient, who works as an NHS health professional, identified the error. She described the experience as "very traumatising" to receive an incorrect diagnosis due to AI and then be told it was a typo.
In another instance, an AI scribe confused a prescribed medication with a different drug of a similar name, an error also caught by the patient. A separate case involved an AI-generated summary letter that failed to include a hospital consultant's instruction for a patient to obtain a repeat prescription for their migraine medication, which could have left the patient without necessary treatment. These examples demonstrate how patients are often the ones identifying these critical mistakes.
Currently, 27 different AI scribes are in use across general practices and hospitals in England. The UK government's ten-year health plan for the NHS in England anticipates that AI scribes will alleviate administrative burdens on staff, freeing up time for patient care. This initiative is central to the NHS's planned transition from an analog to a digital service.
However, Healthwatch England has expressed concern that the Medicines and Healthcare products Regulatory Agency (MHRA) has not classified AI scribes as medical devices. This classification decision means there is no England-wide oversight to ensure the safety and effectiveness of these tools. The MHRA issued guidance on July 29, 2026, clarifying that summarization alone is not sufficient to trigger medical device requirements, and that AVTs must have a "medical purpose" to fall under the definition of a medical device. The MHRA guidance states that AVTs intended solely for transcription, summarizing conversations, drafting letters, or suggesting clinical codes for review are not regulated as medical devices.
Rachel Power, chief executive of the Patients Association, noted that "trust and confidence in this technology depend on good communication and genuine partnership with patients and right now both are missing." A spokesperson for Healthwatch highlighted the "urgent need for clarity over how patients can report and get corrected any mistakes made by AI scribing tools or the professionals that use them."
The rapid adoption of AI scribes has outpaced validation and oversight, according to researchers. Concerns about the accuracy and reliability of AI-generated clinical information persist. Research from September 2025 indicated that while AI scribes offer potential efficiency gains, their deployment without greater scrutiny could compromise patient safety. Another study from November 2025, which analyzed user feedback from a large U.S. hospital system, suggested that AI scribes may introduce patient safety risks due to transcription errors, particularly concerning medication and treatment.
The NHS England published guidance on April 27, 2025, to assist those adopting ambient scribing products. This guidance outlines key considerations for technology leaders implementing these products. It also notes that software products with a 'medical purpose' for individual patients may qualify as a medical device. However, the MHRA's recent guidance appears to overrule an earlier NHS England position that AI scribes performing summarization would likely qualify as medical devices.
Healthwatch England has also found that 81% of people want health professionals to seek explicit consent when AI scribes are used. Furthermore, two-thirds of people indicated that they would feel more comfortable with the technology if human health professionals checked the accuracy of AI-generated summaries. Current NHS guidance states that explicit patient consent is not required for AI scribing tools used for direct patient care, but services should inform patients of their use.