Key points from article :
The UK needs new rules for artificial intelligence (AI) products used in the NHS and wider healthcare, according to an independent commission set up by the Medicines and Healthcare products Regulatory Agency (MHRA). Its report, published on 10 September, makes 44 recommendations and drew on input from more than 12,000 people, including patients and clinicians. In October, the government said it would accept all of them.
The central idea is that AI tools should be monitored for as long as they are in use, rather than approved once and left alone. Unlike a hip replacement or a stethoscope, many AI systems keep changing after approval as they take in new data, and their performance can drift. MHRA chief executive Lawrence Tallon said current medical device rules were written with traditional products in mind. They work for simple AI tools trained to spot known signs on scans, he said, but not for more complex models.
The recommendations include withdrawing approval from AI products that malfunction or become less effective, giving patients the right to know when AI is involved in their care, and powers to penalise developers whose products fall short. The commission also proposed an AI "L plate" scheme, which would let clinicians trial new AI tools under close supervision.
AI is already common in some parts of care. AI note-taking "scribes" are reportedly used by around 40% of UK GPs. Henrietta Hughes, a GP and the Patient Safety Commissioner for England who sat on the commission, said many of her patients were happy for AI to be used, but some chose to opt out, and doctors remained responsible for correcting any mistakes in AI-written notes. Fellow commission member Alastair Denniston, an ophthalmologist, said AI could prove as important to medicine as antibiotics and MRI scans.
The recommendations are not yet law, and a full plan for putting them into practice is not due until spring 2027. Concerns also remain about AI tools making biased decisions because of gaps in the data they learn from, and about chatbots giving incorrect medical advice. Tallon acknowledged that no country has yet found a way of regulating AI in healthcare that fully works.
The report was produced by the National Commission into the Regulation of AI in Healthcare and published by the MHRA.


