UK backs all 44 AI healthcare rules. Now the hard bit
The government has accepted every recommendation from the MHRA's doctor-led AI commission, promising lifelong monitoring of NHS AI tools and faster supervised access. The detail arrives from December.
On 6 October the UK government said it would accept all 44 recommendations from the National Commission into the Regulation of AI in Healthcare, a doctor-led review set up by the medicines and devices regulator, the MHRA. The promise: NHS AI tools watched for their whole working life, not waved through once and forgotten, with promising tools reaching patients sooner under supervision.
"Innovation must never come at the expense of patient safety."
— James Frith, Health Innovation Minister
A rulebook built for scalpels, not software
The Commission was established by the MHRA in September 2025 as an independent advisory body and published its report on 10 September 2026, after a year-long, UK-wide evidence programme involving more than 12,000 people. Its core finding, as pharmaphorum summarised it, is that the current framework was designed for conventional medical devices and is not adequate for AI.
The rules assume a product stays the same after approval. A hip implant does. An AI model that is retrained or quietly swapped onto a new foundation model does not. The Commission's answer is "lifecycle" regulation: proportionate checks before launch, then monitoring for as long as the tool is in use.
What happens next, and when
Unusually for tech policy, the response comes with dates:
- December 2026: the MHRA will publish draft guidance on how to handle changes to AI-enabled medical devices as they evolve.
- 2027: the MHRA will start consulting on how AI devices should be qualified and classified.
- Staged authorisation: the government will explore pathways that let promising AI tools be used in the NHS earlier, under close supervision, while real-world evidence is gathered. Pharmaphorum likened it to "L plates".
- Spring 2027: a full implementation roadmap covering all 44 recommendations.
Alongside that, the MHRA's AI Airlock, its regulatory sandbox for AI medical devices, has opened Phase 3 with three years of extra government funding. A webinar for would-be applicants is set for 22 October at 10:00, with the first wave of innovators due to be picked in November.
MHRA chief executive Lawrence Tallon called the Commission's report "a clear and ambitious blueprint" and said the aim is "making sure patients can benefit from safe, trusted AI more quickly". Professor Alastair Denniston, the Commission's chair, said the response shows the government "has listened, not just to the Commission, but to the more than 12,000 patients" and others who gave evidence. AI Minister Kanishka Narayan said: "AI has huge potential to help doctors deliver better care and improve outcomes for patients."
Ada's take: the right shape, but mostly still a plan
Accepting every recommendation is the easy part. Nothing about how an AI tool is regulated changed on 6 October; what changed is the to-do list. Judge this by the December guidance and the spring roadmap, not the press release.
Still, the direction is sensible. Most patients will meet healthcare AI in two forms. One is the AI scribe, which listens to a consultation and drafts the notes. The MHRA's July guidance said a scribe that only transcribes or summarises is generally not a medical device, but one that suggests diagnoses or guides treatment can be. The other is diagnostic software that reads scans or flags risks, which is squarely regulated. Ongoing monitoring matters most for both, because their errors tend to show up in use, not in a pre-launch test.
One regulator, four health services
This is a genuinely UK-wide story with a devolved twist. The MHRA regulates medical devices across the UK, but healthcare itself is devolved: the NHS in England, Wales and Scotland and Health and Social Care in Northern Ireland decide what they actually buy and use. Health ministers from Wales, Scotland and Northern Ireland all welcomed the response.
Northern Ireland adds a wrinkle. Under the Windsor Framework, EU medical device rules apply to devices placed on the market there, though the MHRA remains the competent authority. How staged authorisation would work there is not yet clear.
To be fair to the sceptics
Supervision costs staff time, which the NHS is short of. The thorniest question, who is to blame when an AI tool gets it wrong, is still being worked on. The Pharmaceutical Journal reported the Commission's warning that responsibility for errors may be transferred onto healthcare professionals and providers; the government has committed only to reviewing how responsibility is split between manufacturers and providers. Tase Oputu, president of the Royal College of Pharmacy, said pharmacy professionals "should remain accountable for decisions within their control" and called for clarity from regulators.
Real-world safety evidence is also thin. The Health Services Safety Investigations Body launched an investigation into AI scribes in hospitals in August because the patient safety implications are "not fully understood"; its report is not expected until summer 2027.
Mind the hype. An AI tool on L plates still needs someone in the passenger seat.
Sources: GOV.UK (MHRA) · pharmaphorum · The Pharmaceutical Journal · GOV.UK: devices in Northern Ireland · Brodies (AI scribes guidance) · Mills & Reeve (HSSIB investigation)