Ministers accept all 44 recommendations to overhaul UK regulation of medical AI

The MHRA will publish draft rules on updating AI devices by December, as its regulatory sandbox opens a third phase devoted to monitoring tools once they reach patients.

The UK government has accepted in full the 44 recommendations of the National Commission into the Regulation of AI in Healthcare, committing the UK to a system in which AI-enabled medical devices are monitored throughout their working life rather than judged largely on a single assessment before they reach the market.

The response, published on Tuesday by the Medicines and Healthcare products Regulatory Agency (MHRA) and ministers, sets out a handful of first steps but leaves most of the detail for later. A full implementation roadmap, assigning timelines and responsibilities for each recommendation, is not due until spring 2027.

The most concrete commitment concerns software that changes after it is approved. The MHRA said it would issue draft guidance by December on how manufacturers should manage modifications to AI devices as they are retrained and improved, a problem the existing devices framework, designed around fixed products, handles awkwardly. Next year the regulator will consult on how AI tools should be qualified and classified as medical devices, the Commission’s first recommendation. Ministers will also explore staged authorisation, allowing promising tools into NHS use earlier under close supervision while real-world evidence accumulates, an approach the Commission compared to L-plates for learner drivers.

The Commission was set up by the MHRA in September 2025 as an independent, non-statutory advisory body. It was chaired by Professor Alastair Denniston, an NHS consultant, with Professor Henrietta Hughes, the Patient Safety Commissioner for England, as deputy. Its report, published on 10 September, drew on a year of evidence gathering involving more than 12,000 patients, clinicians and members of the public, which the regulator describes as the largest such exercise on health technology regulation. It found that public support for AI in healthcare was conditional rather than automatic, and concluded that regulation needed to become more proportionate, lifecycle-based and system-wide.

The report also took up liability, a persistent concern among clinicians. It observed that under existing negligence procedures, claims are disproportionately brought against healthcare professionals and providers, because they carry the clearest duty of care. The government’s response commits to improving access to redress where care falls below expected standards, though Tuesday’s announcement did not say how.

James Frith, the health innovation minister, said the government wanted the NHS to adopt the best new technologies, adding that “innovation must never come at the expense of patient safety.” Hughes said she particularly welcomed the commitments on monitoring AI in practice, patient involvement and routes to redress. “Patients must remain at the heart of how these technologies are developed, regulated and used across healthcare,” she said.

Alongside the response, the MHRA opened applications for the third phase of AI Airlock, its regulatory sandbox for AI medical devices. Phase 3 will concentrate on post-market surveillance, testing with real products how tools can be monitored and managed after deployment. The Airlock was launched in spring 2024; its pilot worked with four companies, Philips, Newton’s Tree, OncoFlow and Automedica, and a second phase, completed this year, covered seven technologies spanning AI clinical note-taking, cancer diagnostics, eye disease detection and obesity treatment support. Earlier Airlock work had already identified the difficulty: adaptive AI strains traditional periodic surveillance, leaving open what should be monitored, how the data is governed and who is responsible for acting on warning signals.

The programme has secured £1.2 million a year from the Department of Health and Social Care for 2026 to 2029. A webinar for prospective applicants is scheduled for 22 October, applications close in November, and the first wave of participants will be selected the same month.

Industry groups welcomed the decision while stressing delivery. Steve Lee of the Association of British HealthTech Industries said there was significant international interest in the UK’s approach and that “building on, and aligning with, international best practice will be key.” Lord Willetts, chair of the Regulatory Innovation Office, pointed to the MHRA’s recent guidance on ambient scribes, the AI tools that draft clinical notes from consultations, as early momentum.

Although the MHRA regulates medical devices across the UK, healthcare delivery is devolved, and the response drew differing reactions from the nations. Health ministers in Wales and Northern Ireland welcomed the recommendations and pledged to work with the regulator. Scotland’s health secretary, Angela Constance, said the Scottish Government shared the Commission’s ambition but would set out its own response in due course, with its AI policy framework for health and social care due later this year.

A newly established programme board will now oversee work across government, the MHRA, NHS bodies and the devolved administrations on nine priority areas identified in the response. The first test of the new approach will come in December, when the regulator’s draft guidance on changing AI devices is published for scrutiny.