The government has accepted all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare, and set out how it intends to deliver them across the UK.
The commission, led by NHS doctors and chaired by Professor Alastair Denniston, published its report on 10 September. Its central conclusion was that the current regulatory approach must evolve for AI technologies, by designing a more proportionate, lifecycle-based framework.
The government confirmed its acceptance on 6 October. The response, it said, commits the UK "to developing a world-leading agile approach to AI device regulation that supports innovation while maintaining patient safety and public trust".
Among the first steps, the Medicines and Healthcare products Regulatory Agency has committed to issue draft guidance by December 2026 on a new approach to managing changes to AI-enabled medical devices as they adapt and improve over time.
The MHRA will also begin next year by consulting on how to qualify and classify AI-enabled devices. That is the commission’s first recommendation.
A further commitment is to explore new staged authorisation pathways, allowing promising AI tools to be used in the NHS earlier under close supervision while further real-world evidence is gathered. A full implementation roadmap, setting out timelines and responsibilities across all 44 recommendations, will be published by spring 2027, with regular progress updates to follow.
Alongside the response, the MHRA opened applications for the third phase of AI Airlock, its regulatory sandbox for AI-enabled medical devices, which has secured three years of additional government funding. A webinar for prospective applicants is scheduled for 22 October, and the first wave of innovators will be selected in November.
"The independent National AI Commission, led by NHS doctors, has given us a clear and ambitious blueprint for how we make the UK the best place to get AI products to patients safely, quickly and in a way that maintains public trust," said Lawrence Tallon, the MHRA’s chief executive.
Denniston said the response showed that government had listened "not just to the Commission, but to the more than 12,000 patients, clinicians and members of the public who shaped our recommendations". A newly established Programme Board will convene to take forward the ten priority areas set out in the response; the published announcement was amended on the day to say there were ten, having previously said nine.
The acceptance rate deserves to be stated plainly. Forty-four recommendations were made and 44 were accepted, which is a hundred per cent, and a hundred per cent is the highest figure available. On the only measure published so far, the framework is complete.
What it will consist of is a separate question, and one the recommendations are themselves about. The first of the 44 asks how AI-enabled devices should be qualified and classified, and that consultation begins next year. Draft guidance on managing changes to those devices arrives in December, which is to say that the rules for altering the category will be published before the category.
This is not a gap. It is the agility. A framework fixed enough to describe would not be able to adapt and improve over time, which is the property the devices are required to have and which the regulator has reasonably adopted for itself. The roadmap setting out timelines and responsibilities across all 44 recommendations is due by spring 2027, at which point the timelines will have a timeline.
Our analysts note that the one component already delivered is the count, and that even it moved: nine priority areas became ten on the day of publication. The sandbox is on its third phase and the Programme Board on its first. The Department of Unnecessary Research therefore nominates the single figure in the system that cannot be revised, and recommends it be adopted as the standard. Forty-four remains 44 whatever the 44 turn out to be.