September 2026

In this issue, our community reacts to the Medicines and Healthcare products Regulatory Agency (MHRA) recommendations on AI in healthcare, and we round up a reading list spanning AI scribes to period-tracking apps. Four paywalled articles are free for newsletter readers for two weeks, including an interview with Alastair Denniston, chair of the commission set up by the MHRA, on the AI recommendations.

The sense from our event advisory committee this month was that the promise of AI sits against a current of public fear, fed by headlines warning of existential risk by the end of the decade.

Our upcoming BMJ Future Health Europe event takes that on, from the "dark side of AI" to the momentum in women's health, and how to build a career in digital health without leaving your day job.

What the new AI recommendations mean for clinicians

The National Commission into the Regulation of AI in Healthcare, set up by the MHRA, published its recommendations for how the UK should regulate AI in healthcare on Thursday. Drawn up after a research and engagement programme with patients, the public, clinicians, industry and regulators, the report proposes a risk based approach: tougher checks for high risk tools such as diagnostic AI, and lighter regulation for lower risk tools such as those handling administrative tasks. Its chair, Alastair Denniston, an ophthalmologist and professor of regulatory science and innovation at the University of Birmingham, set out the thinking to The BMJ: AI medical tools used in "potentially life threatening" scenarios need tighter checks, says head of UK government review. 

Keith Grimes (founder and chief executive of Curistica, former NHS GP, and BMJ Future Health advisory committee member) told BMJ Future Health on Friday that the report has three themes:

  1. The commission proposes regulating what a product does, not the product as a whole. A scribe's transcription would sit outside medical device regulation, while advanced decision support might fall inside it. This rebalances the burden of proof from pre-market to post-market.
  2. It spreads responsibility across the system. Manufacturers would have to specify the controls a clinician's organisation needs in place, and the under-used Yellow Card scheme, used to report safety concerns, would be improved for AI, with reporting designed into the product in use.
  3. It aims to build trust. This includes patients' reasonable expectation that clinicians tell them when AI is used in their care, with an opt-out "where possible or appropriate." Liability remains unclear: the commission asks the Department of Health and Social Care and health organisations to make responsibility, in its words, "clear and appropriate to their respective roles," but says nothing on when or how.

Jools Symons, a member of the BMJ Future Health advisory committee and patient and public involvement manager at the Leeds Institute of Medical Education (LIME), University of Leeds, told BMJ Future Health on Friday that her concern was that the technology develops faster than humans can use it safely: "Can we keep up? I'm not sure the paper reassures me on this issue."

Kim Edwards, a member of a patient and carer community convened by Symons, said a "healthcare professional cannot consent me for a procedure they do not understand." She described a clinician who could not explain a technology when asked. "This was a problem for me, and I would like this report to reassure me that people using AI understand it in the future," she said. Naseem Akram raised the question of how often the recommendations would be reviewed and shared with the public. Joannie Tate asked whether there would be evidence of the report's success, and whether the public would find out if it failed. Cassie Reynolds-Araji said she "would love for AI to be introduced as a member of the [multidisciplinary team] rather than the dominant voice." 

What's your read on the recommendations?

Email me, cschuster-bruce@bmj.com, we feature community responses in future issues. 

Interested in hearing more from Dr Alastair Denniston?He's speaking at BMJ Future Health Europe with Helen Surana, editor in chief at BMJ Innovations. More information about the session can be found below.

29 October 15:15 - 16:00
Evaluating Innovation Across the AI Lifecycle

Healthcare needs new approaches to evaluation that are as agile as the technology itself. Traditional methods alone cannot keep pace, but neither can we afford to compromise on safety, effectiveness, or public trust. In this session, attendees will gain practical insights into rapid evaluation methods, proportionate regulation, and continuous learning approaches that enable healthcare organisations to adopt AI with confidence while ensuring technologies remain safe, effective, and fit for purpose.

Stay current in digital health and AI

We selected these papers from BMJ Group journals to help you keep pace with the issues that arise at the intersection of technology and healthcare. 


Thank you for reading, we look forward to meeting some of you at the Europe event on 29-30 October at The Brewery, London.

Loading