National Commission into the Regulation of AI in Healthcare
The AOP’s summary – and what it means for optometry
What has been published?
On 10 September 2026, the Medicines and Healthcare products Regulatory Agency (MHRA) published the final report of the National Commission into the Regulation of AI in Healthcare.
The Commission was established in September 2025 as an independent body to advise the Government with a new regulatory framework for AI in healthcare. It is chaired by Professor Alastair Denniston, a consultant ophthalmologist, with the Patient Safety Commissioner for England, Professor Henrietta Hughes, as Deputy Chair.
The stated mission of the AI Commission was to “help ensure that patients can benefit from safe and effective AI technologies, support the Government’s ambition to make the NHS the most AI-enabled healthcare system in the world, and establish a globally competitive regulatory environment that attracts investment and supports innovation.”
The report makes 44 recommendations. Its main conclusion is that the current medical device framework is “not fit-for-purpose” for AI. This is because AI products can change over time and perform differently in real-world settings, so a single point-in-time check before they reach the market is not enough. The Commission wants regulation that is proportionate, covers a product’s whole lifecycle, and shares responsibility across the whole health system.
The Commission’s recommendations aim to create a stronger regulatory framework for AI in healthcare, built around three principles: “proportionate life-cycle regulation; system-wide responsibility and safe management; and trust, transparency and predictability.”
Key proposals include a new risk-based system for classifying AI medical devices, tighter monitoring of their performance once in use, clearer accountability for manufacturers and healthcare providers, and better systems for tracking outcomes and reporting safety concerns. The recommendations also emphasise the need for appropriate training for healthcare professionals using AI-enabled tools and clearer routes for patients to seek redress if care falls below expected standards.
The Government has accepted all 44 recommendations and has identified 10 priority areas for delivery over the next year. However, most commitments remain at the exploratory stage, with further detail expected by spring 2027 and no dedicated funding announced. Alongside its response, the MHRA has launched the next phase of its so-called “AI Airlock programme”, which allows regulators and developers to test how AI-enabled medical devices can be safely monitored and managed after they enter clinical use.
The Commission’s full report and the Government’s response are available on GOV.UK.
What do we say?
The AOP is pleased to see the output of the Commission’s report. AI is already beginning to support clinical decisions in optometric practice, particularly in the interpretation of OCT and fundus imaging, and it is likely that AI use will grow. A framework that is proportionate, keeps patients safe and gives practitioners clarity is needed for both patients and the profession.
We particularly welcome the Commission’s recognition that responsibility for AI should not simply fall on the frontline clinician. Making sure that contracts set out who is responsible for risks that result from AI use is important for the safe use of this technology. Too often practitioners are expected to take complete clinical responsibility for tools they have limited visibility of and little control over. We have previously shared our concerns about the risk of the black box, a device that is beyond the understanding of many of those using it. In this instance it is almost impossible for an employed optometrist to decide if technology is appropriate and they must have sufficient support to be able to exercise their professional judgement. For the profession, there is an important balance to be struck as the risk is two-fold.
If we defer all decision making to the device, optometrists could become little more than technicians, but take on too much responsibility and we will be exposing ourselves to risks that we cannot fully understand.
Alongside the positives, many of the recommendations place new obligations on providers, including showing readiness, delivering the risk controls a manufacturer specifies, training staff and recording device versions in the patient record. Much of the report has been written with NHS trusts in mind. Our view is that community optical practices, many of whom are small independent businesses, should have proportionate expectations on them and practical support to prevent disadvantage when adopting technology that may help improve patient care. We will be pressing for primary eye care to be explicitly considered in the design of the AI readiness toolbox and the national guidance that will follow.
We also note the recommendation that professional regulators help build AI capability across education and CPD. We will engage with the GOC and education providers to make sure any new expectations are realistic, clearly defined and supported, and don’t simply add requirements without the training to meet them.
In its evidence to the Commission in January 2026, the GOC said the existing law is unfit for AI, called for clear lines between the MHRA and the professional regulators, and stressed that registrants remain professionally accountable for decisions made with AI support. As we have said above, we agree with this point, but recognise there is a difficult balance to be found.
The GOC also committed to developing guidance on the safe use of AI in eye care once the Commission had reported. The Commission has now reported and the Government has responded, but the GOC has yet to set out its own response. The GOC’s Standards Committee has discussed bringing forward draft AI guidance, and the GOC is helping the Professional Standards Authority and MHRA working group develop shared AI principles for health professional regulators, due by January 2027. The GOC 2026 registrant survey showed that three-quarters of registrants are concerned about AI making errors and about who is legally responsible when things go wrong, and six in 10 rate their understanding of AI as poor. We will monitor the GOC Council meetings to see if the GOC brings its guidance to Council early in 2027 as planned.
We hope the GOC will build on those principles and work with us so that it is practical for practitioners and consistent with the responsibilities being clarified nationally, and to make sure its review of business standards addresses who is responsible for the AI tools that practices buy.
We also welcome the Government’s decision to accept every recommendation made by the Commission. However, we note that accepting the recommendations is not the same as delivering them, and much of the response is framed as options to explore or consider, with little detail expected before spring 2027.
In particular, we are concerned that the response makes no reference to primary or community care. Neither does it name the GOC among the bodies taking the work forward. We will press for community optical practice to be represented in the Programme Board’s working groups, particularly on shared responsibilities such as the AI readiness toolbox and workforce capability.
We note that the new working group on responsibilities will clarify, but not change, where responsibility lies. For optometrists, who are individually accountable, that clarity must be practical, and it must be reflected in the contracts that practices sign with manufacturers.
What happens next?
The timeline for future work is as follows:
- November 2026: the MHRA is due to select the first group of developers for AI Airlock Phase 3
- December 2026: the MHRA will publish draft guidance, which sets out in advance how an AI product may change after authorisation, for public comment (Recommendation 6)
- January 2027: shared AI principles for health professional regulators are due from a Professional Standards Authority and MHRA working group that includes the GOC
- Spring 2027: the MHRA consults on how AI products are qualified and classified as medical devices, to be followed by secondary legislation (Recommendation 1). This is likely to affect many of the image-analysis tools used in practice today, which are self-declared as Class I devices
- Spring 2027: the Government publishes an implementation plan and roadmap, with annual progress updates after that.
When the MHRA publishes draft guidance, we will respond. We will also engage with the DHSC and the GOC as this work develops, respond to the consultations as they open, and update members on the practical implications for practice.
Practical guidance for members on the use of AI in practice is available from our AI and technology hub.