Qurexa Logo
Qurexa

Trust. Care. Deliver.

Back to Blog & Insights
Technology

All 44 accepted: how Britain plans to regulate AI in healthcare

Qurexa Editorial Team11 October 20268 min read 0 0
All 44 accepted: how Britain plans to regulate AI in healthcare

What happened

On 6 October 2026, the government said it would accept all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare. All 44. Not most of them, and not in principle. In full. The Commission was set up by the Medicines and Healthcare products Regulatory Agency (MHRA) in September 2025 as an independent advisory body. It published its report on 10 September 2026. Less than a month later, the government responded by accepting the lot. A new Programme Board has been created to oversee work on ten priority areas. A full implementation roadmap, setting out timelines and who is responsible for each of the 44 recommendations, is due to be published by spring 2027. Some specific commitments have firm dates already. The MHRA will publish draft guidance by December 2026 on how to manage changes to AI-enabled devices as they adapt over time. In early 2027 it will start consulting on how to qualify and classify AI-enabled devices, which was the Commission's first recommendation. The MHRA is also opening applications for Phase 3 of AI Airlock. AI Airlock is its regulatory sandbox: a controlled space where AI-enabled medical devices can be tested with the regulator closely involved. A webinar for prospective applicants is set for 22 October 2026 at 10:00, and the first wave of Phase 3 innovators will be selected in November 2026. AI Airlock has secured three years of additional government funding. The government is also exploring staged authorisation pathways. That would let promising AI tools be used in the NHS earlier, under close supervision, while more real-world evidence is gathered.

Why this matters

AI is already in the NHS. That is the part of this story that gets missed. Software helps read scans, flag suspicious chest X-rays, sort referrals and spot patients at risk of getting worse. This is not a future question. It is a question about tools that are in use now. The awkward problem is that medical device rules were mostly written for things that do not change. A hip replacement is the same hip replacement in five years. A pacemaker behaves the way it was approved to behave. You can test it once, approve it, and trust that approval. AI is not like that. It can be updated, retrained and tuned. It can perform brilliantly on the population it was trained on and much worse on a different one. A tool tested mainly on one group of patients may quietly be less accurate for another. So a one-off approval is the wrong shape for the problem. The Commission's central idea is to replace it with lifecycle regulation: assess and monitor these devices across their whole working life, not just at the start. For patients, this is really a question about who is watching. If a scan-reading tool starts making more mistakes in month fourteen than it did in month one, somebody needs to notice. Lifecycle regulation is the mechanism for noticing. James Frith MP, the Health Innovation Minister, framed the trade-off directly: "innovation must never come at the expense of patient safety."

What the evidence actually says

The strongest thing about this process is how much listening went into it, and that deserves more attention than the headline number of recommendations. More than 12,000 patients, clinicians and members of the public took part in a year-long, UK-wide evidence-gathering programme. That is a serious sample for a regulatory review. It means the recommendations are not purely a technical exercise written by specialists for specialists. The Commission's recommendations cover more than approval rules. They include clear rules for qualifying and classifying AI-enabled devices, stronger patient engagement in decisions about AI, better access to redress when standards of care fall short, and fair sharing of AI's benefits across all communities, including underserved groups. That last point is worth pausing on. It is an admission built into the recommendations that AI in healthcare could widen gaps as easily as close them, if the tools work less well for some groups than others. Professor Alastair Denniston, who chaired the Commission, said the government's response "shows that government has listened" to the Commission and the public. Now the honest caveats, because accepting recommendations is not the same as acting on them. The detailed roadmap does not exist yet. It is due by spring 2027. Until then, we know the direction but not the timings for most of the 44 items. The consultation on qualifying and classifying devices has not started. It begins in early 2027. And this is a devolved area in part. Mabon ap Gwynfor, the Welsh Cabinet Minister for Health and Care, welcomed the recommendations and said Wales will work closely with the MHRA. Robbie Butler, Northern Ireland's Health Minister, said Northern Ireland's perspectives were reflected in the work. Angela Constance, the Scottish Health Secretary, said Scotland shares the ambition but will set out its own response later. So the UK-wide picture is not identical everywhere. Lord David Willetts, Chair of the Regulatory Innovation Office, welcomed the response and called for the recommendations to be turned into "real-world action". That is the fair test, and it has not happened yet.

Practical advice

For most people, the useful response here is not technical. It is knowing what you are allowed to ask. You can ask whether AI was involved in your care. If a scan, referral or triage decision involved software, you are entitled to ask about it. Professor Henrietta Hughes OBE, the Commission's Deputy Chair and Patient Safety Commissioner for England, said patients "must remain at the heart" of how AI is developed, regulated and used. Asking is part of that. Remember that a clinician remains responsible. AI tools in the NHS support decisions; they do not replace the professional making them. If something about a decision does not make sense to you, the person to raise it with is the clinician, not the software. Be careful with the difference between NHS tools and consumer apps. An AI symptom-checker you downloaded yourself has not been through the process described here. A regulated medical device has. They are not the same thing, even if they look similar on a screen. If you develop or sell AI-enabled medical devices, the dates matter now. Draft guidance on managing changes to AI devices is due in December 2026. The AI Airlock Phase 3 webinar is on 22 October 2026, with the first innovators selected in November. Consultation on qualification and classification starts in early 2027. If you want to shape the rules rather than react to them, these are the openings. If you work in the NHS, the direction of travel is towards monitoring tools in use, not just at purchase. Sir Andrew Goddard, who chairs the AI Airlock Governance Board, said the next phase will focus on how AI products can be monitored once they are in use. And if you use a health app that worries you, you can report problems with medical devices to the MHRA through its Yellow Card scheme.

What to know

The government has accepted all 44 recommendations of the National Commission into the Regulation of AI in Healthcare, which was set up by the MHRA in September 2025 and reported on 10 September 2026. More than 12,000 patients, clinicians and members of the public contributed to the evidence. The core shift is from one-off approval to lifecycle regulation: assessing and monitoring AI-enabled medical devices throughout their working life, because they change over time in ways traditional devices do not. Key dates: draft MHRA guidance on managing changes to AI devices by December 2026; AI Airlock Phase 3 applicant webinar on 22 October 2026 with first innovators chosen in November; consultation on qualifying and classifying AI devices from early 2027; full implementation roadmap by spring 2027. The fair verdict: a genuinely thorough piece of work, accepted unusually completely, with the hard part still ahead. Scotland will respond separately, and the detailed timings for most of the 44 recommendations are not public yet. Sources: MHRA (GOV.UK), 'Government backs recommendations of NHS doctors-led AI Commission', 6 October 2026, https://www.gov.uk/government/news/government-backs-recommendations-of-nhs-doctors-led-ai-commission; MHRA, 'National Commission into the Regulation of AI in Healthcare', report published 10 September 2026, https://www.gov.uk/government/groups/national-commission-into-the-regulation-of-ai-in-healthcare; MHRA, 'AI Airlock: regulatory sandbox for AI as a medical device', https://www.gov.uk/government/collections/ai-airlock-the-regulatory-sandbox-for-aiamd. This article is for general information and does not replace advice from a doctor, pharmacist or other qualified healthcare professional. If you have questions about how a decision about your care was reached, including the role of any software, raise them with your own clinical team.

#artificial intelligence#MHRA#medical devices#NHS#regulation#patient safety

Related articles