The UK will keep monitoring AI medical tools as long as they are in use, not just before approval. AI-based medical devices will need constant monitoring even after approval, rather than a single pre-approval safety check, according to the head of the UK’s medicines regulator. For patients, that means ongoing safety checks, clearer information about when AI is used in your care, and better routes to complain when something goes wrong.
What is the UK’s new rule on continuous monitoring of AI in healthcare?
The UK government has committed to continuously monitoring AI in healthcare throughout its working life. Regulators will track how these tools perform in real healthcare settings, instead of relying heavily on one-off assessments. The Medicines and Healthcare products Regulatory Agency (MHRA) will lead it.
A quick correction to the headline wording, because it matters. This isn’t a finished law yet. On 6 October 2026 the government accepted all 44 recommendations from the National Commission into the Regulation of AI in Healthcare. The detailed rules still have to be written, consulted on and rolled out.
Why a one-off approval check isn’t enough for AI
Think about how a normal medical device gets approved. A scanner or a blood pressure cuff is tested, certified and sold. It behaves the same on day 1,000 as on day one.
AI doesn’t always work that way.
MHRA chief executive Lawrence Tallon described the new era as one of non-deterministic algorithms, meaning AI systems that evolve and adapt while doctors are using them. A tool that was accurate in one hospital can behave differently in another. Performance can vary with the setting, workflows, people and surrounding organisations. A single check at the start can’t catch that.
So the answer is to keep watching. Not once. Continuously.
What changes for patients
Four things stand out from the government’s patient explainer and the wider announcement.
- Safety checks that don’t stop at launch: There will be stronger monitoring to spot and act on changes in safety or performance once AI is in use.
- More openness about AI: Patient involvement will be strengthened, including greater transparency about when AI is used in healthcare.
- A route to redress: The same package promises clearer routes to redress when standards of care fall below expectations.
- A bigger say: Patients will get better information about how AI is used and more chances to influence how it is regulated.
Your clinician stays in charge. The MHRA says clinicians’ professional judgement stays central, with AI supporting their expertise and decision-making rather than replacing it.
How will the monitoring actually work?
Nobody has the full answer yet, and anyone claiming otherwise is guessing. But there’s a testing ground running now.
The MHRA has opened phase 3 of its AI Airlock, a regulatory sandbox. It will work with developers, regulators and healthcare partners to test how AI devices can be safely monitored after deployment, and the findings will feed directly into future MHRA guidance. The first selections are due in November.
Responsibility also gets spread around. The commission’s framework shares it across manufacturers, providers, clinicians and regulators, so hospitals that buy AI tools will carry new duties too. That’s a sensible shift. When something goes wrong with an AI tool, “the vendor’s problem” is no longer a complete answer.
Timeline: what happens and when
| Date | What happens |
| September 2025 | MHRA launches the National Commission into the Regulation of AI in Healthcare |
| 10 September 2026 | Commission publishes its 44 recommendations |
| 6 October 2026 | Government accepts all 44 recommendations |
| November 2026 | First AI Airlock phase 3 selections |
| December 2026 | MHRA publishes draft guidance on managing changes to AI-enabled devices |
| Early 2027 | Consultation on how to classify and qualify AI-enabled devices |
| Spring 2027 | Full implementation roadmap published |
Sources for these dates are linked at the end. The December draft guidance, the early-2027 consultation and the spring 2027 roadmap are the three milestones to watch.
Will AI replace my doctor?
No. Nothing in the announcement points that way. The stated goal is AI that supports clinical decisions, with the clinician accountable. The aim is also speed: the Commission wants a framework that is safe, fast and trusted, so useful tools reach patients sooner without cutting safety corners.
What can you do as a patient right now?
Ask. If a clinician mentions an AI tool in your care, it’s reasonable to ask what it does, what it was trained to do and who checks its results. Today you may get a vague answer. The new transparency commitments are meant to make that conversation easier.
You can also follow the MHRA’s guidance pages, since the public will get chances to comment as the rules take shape. The Commission gathered evidence from more than 12,000 people over a year, so public input clearly shaped this.
FAQ
Is continuous monitoring of AI in healthcare law in the UK yet?
Not yet. The government has accepted the recommendations, but guidance, consultations and a full roadmap are still to come through 2026 and 2027.
Who will monitor AI medical devices in the UK?
The MHRA leads regulation. Manufacturers, healthcare providers and clinicians also share responsibility for safe use.
Does this affect AI tools I use myself, like health chatbots or apps?
The announcement covers AI-enabled medical devices used in healthcare. Whether a particular consumer app counts depends on whether it is classed as a medical device, and the classification consultation in 2027 is meant to clarify that.
What is the MHRA AI Airlock?
It’s a regulatory sandbox where developers, regulators and healthcare organisations test how AI medical devices can be introduced and monitored safely. Phase 3 focuses on monitoring after deployment.
Will I be told when AI is used in my care?
Greater transparency about AI use is part of the package. The exact format hasn’t been set, so expect detail to come through the guidance.
