AI-powered transcription tools used across the NHS are generating dangerous errors in patient records, including misdiagnoses and incorrect medication information, according to warnings from the statutory patient safety watchdog Healthwatch England.
One patient was incorrectly told she had demyelination – serious nerve damage linked to multiple sclerosis – in her AI-generated consultation summary. The error was only corrected after the patient, herself an NHS worker, questioned the AI’s interpretation of her MRI results. “This was eventually corrected but was a very traumatising experience to be given an incorrect diagnosis because of AI and then be told it’s a typo,” she said.
Rapid Rollout Raises Safety Concerns
The government’s 10-year health plan for the NHS in England positions AI scribes as key to “liberating staff from their current burden of bureaucracy and administration.” Currently, GPs and hospital doctors in England are using 27 different AI scribe platforms. However, Healthwatch England has documented multiple instances where patients identified errors that healthcare professionals missed.
In another case, an AI scribe confused a prescribed medication with a similarly-named drug. Separately, an AI-generated summary failed to include a consultant’s instruction for a patient to obtain repeat migraine prescriptions from their GP – potentially leaving them without essential medication.
Lack of Oversight and Regulation
Healthwatch warns that without proper oversight, these inaccuracies could persist in patients’ medical records indefinitely. The organisation has called for urgent clarity on how patients can report and correct mistakes made by AI tools.
More concerning, the Medicines and Healthcare products Regulatory Agency (MHRA) has decided not to classify AI scribes as medical devices, meaning there is no England-wide safety oversight for these tools. “Healthcare has never been error-free,” said a Healthwatch spokesperson, “but our findings show the urgent need for clarity over how patients can report and get corrected any mistakes.”
Medical Professionals Sound the Alarm
Dr Shier Ziser Dawood, a London GP, previously warned in the British Journal of General Practice that AI scribes may prove “a double-edged sword” for family doctors. She detailed how an AI tool incorrectly recorded that she had told a patient to “continue their Prozac” – a drug she had neither prescribed nor discussed with the patient. This type of error, known as “hallucination,” occurs when AI refers to information never mentioned during consultations.
Dr Charlotte Blease, an AI healthcare expert at Uppsala University, acknowledges that while doctors make mistakes without AI assistance, the technology introduces new risks. Her research found that ambient AI records are more prone to errors during consultations involving multiple people, patients with complex medical histories, or those for whom English is not their first language. Despite this, over half of 1,003 UK GPs surveyed believed their AI-generated records were more accurate than self-written notes.
Time Savings Questioned
Contrary to NHS assumptions that AI scribes will reduce administrative burden, Dr Dawood argues that doctors must now spend additional time reviewing transcripts for accuracy. With NHS bosses expecting GPs to see two additional patients daily due to perceived efficiency gains, the reality may be quite different.
Rachel Power, chief executive of the Patients Association, emphasised that trust in AI technology depends on “good communication and genuine partnership with patients – right now both are missing.” The rapid deployment of these tools without adequate safeguards raises serious questions about patient safety and the future of digital healthcare implementation.
Why it Matters
The NHS’s ambitious digital transformation risks compromising patient safety through untested AI technologies deployed at scale. Without mandatory regulatory oversight or standardized error-correction protocols, inaccurate AI-generated medical records could lead to misdiagnoses, inappropriate treatments, and potentially life-threatening delays in care. Patients are currently left to identify errors themselves – an unacceptable burden that shifts responsibility away from healthcare providers. As the NHS continues its rapid AI rollout, these incidents highlight the critical need for robust safety frameworks, transparent reporting mechanisms, and genuine patient involvement in digital health decisions. The stakes are too high to treat patient records as experimental territory.
