The UK government has issued its response to the recommendations made by the National Commission into the Regulation of AI in Healthcare, choosing to accept all 44 findings, and highlight an upcoming implementation plan and roadmap to be published by Spring 2027.
The commission published its recommendations for a future regulatory framework in September, finding: “Current approaches were largely designed for products that are more static and easier to reliably assess at a single point in time. AI-enabled products may iterate rapidly, perform differently in different settings and depend on the data, workflows, people and organisations around them.”
James Frith, parliamentary under-secretary of state for health innovation, provides a ministerial foreword to the government response, noting: “This Government agrees that the UK needs a new approach to AI product regulation in healthcare that is proportionate, risk and lifecycle based and takes a system wide approach to overseeing safe use. This is essential for shifting the narrative on AI adoption away from highlighting the potential benefits of AI in healthcare to demonstrating the benefits in real world settings, enabling successful innovation to be adopted at scale, and building the trust that is fundamental for the use of these pioneering products.”
The government therefore sets out a series of priority areas to be focused on over the next year, starting with modernising the regulatory framework for AI-enabled medical devices to include clearer guidance on when a product qualifies as a medical device, an “improved classification approach”, and clearer guidance on how intended purpose takes into account design and functionality considerations.
Looking to equitable access and outcomes, the government will move to promote approaches that consider equity throughout the lifecycle of AI-enabled technologies, and shift toward stronger post-market assurance with new surveillance mechanisms and improved mechanisms for monitoring safety and performance changes in real-world settings. “Novel” staged authorisation pathways will be explored to allow earlier access while generating real-world evidence, and the use of regulatory sandboxes will be expanded to test new technologies and regulatory approaches, it adds.
Moving on to responsibility, the government shares plans for the MHRA to develop guidance outlining how manufacturers should identify and communicate the operational conditions required for safe deployment and use, involving the provision of practical information and illustrative examples of risk controls. The DHSC and NHSE will work on an AI readiness toolbox to support healthcare organisations in preparing for AI-enabled technologies, and a working group formed of the DHSC, MHRA, CQC and equivalent organisations will focus on strengthening mechanisms for reporting, monitoring, and learning from real-world use.
Trust, transparency, and predictability will also reportedly be tackled, with the MHRA and its partners to increase patient and public engagement in regulatory decision-making, offer clearer guidance and opportunities for early engagement for developers or manufacturers, and explore new approaches to communicating information about AI technologies and safety concerns. The DHSC will work with healthcare partners to improve transparency and consistency in informing patients about the use of AI-enabled technologies in their care.
A cross-system programme board is to be established with representation from key partner organisations to help ensure effective delivery of these plans, the government notes. This board will meet quarterly to check progress and delivery against objectives, as well as to assess emerging risks. An implementation plan and roadmap will be published by Spring 2027, it continues, offering further detail on roles, responsibilities, and key deliverables.
The MHRA also plans to launch a consultation on its new approach to classification and qualification for software and AI-enabled devices by Spring 2027, the government indicates, before using secondary legislation to introduce an updated definition, a classification system that “better reflects a proportionate approach to the benefits and risks” of devices and software, and an updated definition of intended purpose.
Other accepted recommendations include the introduction of a public-facing database or tool for members of the public to search for information on adverse incidents relating to medical devices, the preparation of guidance on cyber security expectations for software and AI-enabled devices, and the exploration of opportunities to automate low burden reporting requirements by embedding reporting into systems like EPRs. A coordinated approach is promised to improving AI literacy across the workforce, aiming to support a shared baseline of general AI capabilities alongside learning tailored to roles and responsibilities.
Wider trend: Health AI
In its latest meeting, the NHS England board raised concerns regarding the lack of a single clear vision for the role of AI in the NHS, and the potential for this to lead to delays and duplication. The board moved on to emphasise the importance of developing proportionate and risk-based arrangements “at pace” that can enable safe innovation and continuing assurance throughout the lifecycle of AI systems.
The Health Service Executive (HSE) for Ireland has published an AI implementation framework outlining a five phase approach for AI projects from identification and prioritisation through to deployment and ongoing monitoring. It is designed to support with the implementation of the government’s AI for Care strategy, published earlier this year. The first of the five phases focuses on opportunity identification and prioritisation, assessing whether opportunities are suitable use cases for AI.
Applications have opened for the sixth cohort of the NHS Fellowship in Clinical AI, offering applicants the chance to be matched with existing clinical AI projects within the NHS to learn about the safe deployment and evaluation of AI in clinical workflows. The fellowship is designed to run for 12 months, with successful applicants receiving salary cover for two days (15 hours) per week of time to commit to their study. Partial remote working and flexible hours may be available depending on the project.
