The publication of the National Commission into the Regulation of AI in Healthcare – Call for Evidence Summary of Findings marks an important milestone for the UK’s ambitions to become a global leader in safe, effective AI-enabled healthcare. Drawing on over 700 responses, including our own, Health Innovation Network South London CEO Dr Rishi Das-Gupta and Director of Digital and Transformation Dr Amanda Begley, discuss how the report highlights both strong consensus across the sector – and several areas where further clarity and action are needed.
Here in south London we have a long record of focusing on the practical benefits of AI in healthcare – from the first roundtable on ambient voice technology with national stakeholders helping coin the term AVT, to the recent report into AI and productivity in the NHS workforce and supporting four Trusts in south west London is the largest procurement and role out of AVT in England. When we saw that there was a national commission looking into the regulation of AI in healthcare we were keen to share our experiences. So what do we think of the findings?
Regulation must be relevant for iterative and adaptive AI systems
Stakeholders reported that the current framework is designed for more static medical devices, and not well suited to iterative and adaptive AI systems. This came with strong support for enhancing post-market surveillance and improving coordination. These were key points mentioned in our submission but we went further in suggesting how this could work, using the example of NICE’s evidence standards for digital health technologies, which provides a strong foundation for categorising innovation types and aligning them with proportionate evidence requirements. A similar tiered model designed specifically for AI, especially adaptive AI, would help ensure regulatory expectations are clear, consistent and achievable.
Strong agreement on the need for better evaluation and evidence
A major theme in the findings is the need for clearer standards and mechanisms for evaluating AI technologies, both before and after deployment.
Our submission strongly reinforces this, calling for national platforms to independently test and validate algorithms, alongside shared evaluation methodologies and better training for clinicians and commissioners. The report similarly highlights challenges around evidence generation, transparency, and consistency in evaluation approaches across the system.
Where our response goes further is in proposing concrete infrastructure solutions – such as independent testing environments and benchmark datasets – to operationalise this vision. The Commission identifies the problem; our response offers a pathway to solving it.
Post-market surveillance: from passive to proactive
There is also strong alignment on post-market monitoring. The Commission recognises that risks often emerge only once AI systems are deployed in real-world settings, and that existing mechanisms may be insufficient.
Our submission expands on this by proposing a structured, data-driven model for surveillance – including tracking usage patterns, clinical outcomes, and anticipated failure modes. We also stress the importance of transparency and standardised national guidance, so NHS organisations can consistently monitor AI performance.
This represents a clear commonality, where both the report and our response advocate for a shift from passive incident reporting to lifecycle-based oversight.
Clarity of responsibility: consensus with a sharper edge
The Commission notes ongoing complexity around accountability and responsibility across the AI lifecycle. Our submission aligns with this concern but takes a more prescriptive stance: the biggest risk is not shared responsibility itself, but poorly defined responsibility. We argue for clearly delineated roles across manufacturers, NHS England, providers, and clinicians – supported by governance such as Clinical Safety Officers.
This reflects a subtle but important difference. While the Commission highlights the issue, our response emphasises the need for explicit, non-overlapping accountability to prevent safety gaps.
Key gaps: health economics and real-world value
One notable difference is the relative absence in the Commission’s summary of detailed discussion around health economic impact and real-world value.
Our submission highlights this as a critical gap in the current regulatory landscape, arguing that AI assessment should include cost-effectiveness and system-wide impact. Without this, the NHS risks approving technologies that are technically sound but fail to deliver meaningful value at scale.
Looking ahead: from insight to implementation
Overall, the Commission’s findings strongly validate many of the priorities outlined in our submission. It plans to consider the findings as it develops and refines the recommendations it will make to the MHRA, health system, and wider government.
At the same time we will continue to offer up our real-world experience to ensure any recommendations will be usable in the real world applications and deliver positive impacts to patients and the NHS.