The National Commission into the Regulation of AI in Healthcare has offered early insight into outcomes from a call for evidence attracting more than 770 responses, indicating strong support for regulatory reform and the need for further development in areas such as post-market surveillance and ongoing monitoring.
Henrietta Hughes, patient safety commissioner and deputy chair of the commission, presented the findings, suggesting the response to the open call for evidence demonstrates “how strongly patients and the public want to be included as partners in the regulation of AI in healthcare”. The overarching message is clear, Hughes went on: trust is central, extending to the technologies themselves, their use, and how they are governed.
“There is strong support for reform of the current regulatory approach,” Hughes explained. “People want meaningful change, but not a complete overhaul. This reflects confidence in parts of the existing system, alongside a need to strengthen and adapt it as technologies evolve.”
Concerns from the call for evidence raised the question of how AI medical devices are monitored once in use, as well as uncertainty around liability, according to Hughes, who points to the importance of ongoing monitoring and transparency.
Wider engagement is now underway, with the MHRA working closely with patients and communities in partnership with National Voices, and in-depth public discussion sessions delivered by The Health Foundation with Ipsos MORI. The MHRA has also led sector roundtables with over 30 organisations represented, including 117 clinicians and representatives from across the health system.
An open “Ask Me Anything” webinar will be held on 20 May, giving patients and the public the chance to hear directly from the commission’s leadership. Henrietta Hughes will be joined by the chair of the National AI Commission and the chief executive of the MHRA to share more about current focuses and respond to questions. To register, or to submit a question ahead of time, please click here.
The commission’s recommendations are expected to be published over summer 2026.
Wider trend: AI in health and care
The Medicines and Healthcare products Regulatory Agency has opened an offering to provide regulatory advice meetings on medical devices and in vitro diagnostic devices for manufacturers, at a cost of £987 for one hour. During the one hour meeting, the person or organisation requesting support is to begin with a short presentation outlining their questions and the “issues and controversies” surrounding them. The service aims to provide regulatory advice relating to medical devices, particularly if the application of existing regulatory guidance is not straightforward.
UK government-backed Sovereign AI has announced the launch of its new £500 million fund, designed to give early-stage AI companies in the UK access to the funds, compute, and strategic assets needed to scale rapidly and compete on a global scale. The announcement offered an insight into what was on offer for UK AI providers, such as access to AI supercomputers with up to one million GPU hours per startup, early-stage investment of up to £20 million, and strategic assets to support with the creation of AI datasets and autonomous lab infrastructure. £282 million is also being dedicated to support AI startups with research and development, with a funding call to be announced for the creation of new datasets and other assets.
A US National Institutes of Health-supported study has developed an AI algorithm trained on EHR data to predict rare disease, with plans to scale over time to suggest when disease may appear, and how patients will respond to treatment. The WEakly Supervised Transformer (WEST) algorithm is reportedly capable of using “noisy”, incomplete, inaccurate, or non-informative data from EHRs to predict whether a patient is likely to have a specific rare condition. The algorithm was initially tested using EHR data from patients at risk for two rare lung diseases: pulmonary hypertension and severe asthma, achieving the highest rated predictive performance among all baseline models in identifying those diagnosed by clinicians.
