UK needs new laws for AI in healthcare, says watchdog

The Medicines and Healthcare Products Regulatory Agency (MHRA) has issued 44 recommendations aimed at modernising the United Kingdom’s regulatory framework for artificial‑intelligence (AI) tools as they become a routine part of National Health Service (NHS) care. Chief regulator Lawrence Tallon told the BBC that patients will soon encounter AI “as part of the way that normal NHS healthcare is delivered” and that the system must preserve public trust. The guidance, produced by an independent commission that consulted more than 12,000 patients, clinicians and other stakeholders, calls for continuous monitoring of AI products, the ability to withdraw approval if a tool drifts or malfunctions, mandatory disclosure to patients when AI is involved in their treatment, and the power to penalise developers that fail to meet standards. It also proposes an “AI L‑plate” scheme to allow new models to be trialled under close supervision, reflecting the agency’s view that existing medical‑device rules—originally designed for static items such as hip replacements—are ill‑suited to software that learns and evolves after market authorisation.

A key consequence of the MHRA’s proposals is the need to balance the promise of AI‑driven efficiency with the risk of eroding patient confidence. Current AI applications in the NHS are largely administrative, such as large‑language‑model (LLM)‑powered “scribe” systems that transcribe consultations for roughly 40 % of UK general practitioners. Yet a University of Edinburgh study found that patients may withhold sensitive information, including substance‑abuse histories, if they know an algorithm is processing the conversation. Professor Henrietta Hughes, a GP on the commission, noted that while many patients welcome AI assistance, some explicitly opt out, preferring not to “talk to a robot.” She stressed that any errors made by scribes remain the clinician’s responsibility to correct. The report therefore underscores the importance of transparent patient consent and robust oversight to prevent bias‑driven misdiagnoses or erroneous advice from AI chatbots, concerns echoed by other experts who warn that AI models can inherit and amplify flaws in the data used to train them.

The broader implications of the MHRA’s recommendations point to a transformative, yet cautious, integration of AI across the health sector. Ophthalmologist Professor Alastair Denniston hailed AI as an “exceptional opportunity” comparable to historic breakthroughs such as antibiotics and magnetic‑resonance imaging, while industry leaders like Arm’s René Haas have even suggested AI could deliver a cure for cancer within a lifetime. Nonetheless, the agency acknowledges that regulation of adaptive AI is a global challenge with no single country having “cracked” the problem. By instituting continuous post‑approval surveillance, mandatory patient disclosures and enforceable penalties, the MHRA aims to create a regulatory environment that can accommodate rapid technological change while safeguarding patient safety and trust, setting a precedent that other nations may follow as AI becomes ever more embedded in clinical decision‑making and medical research.

⚡ Effects Interpreter