MHRA proposes 44 new rules to regulate AI in UK healthcare

The Medicines and Healthcare Products Regulatory Agency has published a set of 44 recommendations aimed at updating the regulatory framework for artificial intelligence products used within the National Health Service and other healthcare settings in the United Kingdom. The agency, which is responsible for regulating medical devices and licensing treatment drugs, argues that current laws are insufficient for the evolving nature of AI technology. These new guidelines are designed to ensure that as AI becomes a routine part of medical care, patients can maintain their trust and confidence in the healthcare system.

Lawrence Tallon, the chief of the MHRA, stated that the agency expects patients to increasingly view AI as an integral component of standard NHS healthcare delivery. He emphasised that this integration must occur in a manner that preserves public confidence. The recommendations were compiled by an independent commission that gathered input from more than 12,000 individuals, including patients and clinicians. Key proposals include the continuous monitoring of AI products, with the power to remove them from regulatory approval if they malfunction or lose effectiveness over time. Additionally, the plan grants patients the right to know whether AI is involved in their care and provides easy access to information about the specific products used.

The regulatory body also proposes the power to penalise developers if their AI products fail to meet required standards. A new system, described as an AI L plate, would facilitate the trialling of new AI models by healthcare professionals under close supervision. Tallon noted that the existing medical devices regulatory framework largely dates from an era focused on static items such as hip replacements, stethoscopes, and plasters. While current guidelines may be adequate for simple AI products trained to identify known symptoms on scans, they are ill-suited for more complex models. Unlike traditional medical products, AI systems continue to change after authorisation, learning, adapting, and drifting as new data is processed.

Tallon acknowledged that regulating AI is a global challenge, stating that no single country or regulatory framework has yet fully solved the issue. In the United Kingdom, AI note-takers powered by large language models are reportedly used by 40 per cent of general practitioners to record consultations and generate reports. However, a recent study from the University of Edinburgh suggested that patients may be less likely to share sensitive personal information, such as substance abuse history, if they know the conversation is being processed by AI. Professor Henrietta Hughes, a general practitioner involved in the report commission, noted that while many patients are comfortable with AI use, some prefer to opt out, stating they do not wish to talk to a robot. She added that while AI scribes can make mistakes, it remains the responsibility of doctors to correct them.

Although AI is currently used primarily for administrative tasks and symptom spotting under human oversight, there is significant excitement regarding its potential in medical research. Professor Alastair Denniston, an ophthalmologist who contributed to the report, described the technology as an exceptional opportunity for healthcare, potentially ranking alongside major advancements like antibiotics and MRI. Despite this optimism, concerns remain about AI models making incorrect decisions due to biases in training data and the risk of inaccurate medical advice from AI chatbots.