The commitment was made as the Government accepted all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare, published on 10 September.
Health innovation minister James Frith said: ‘This will be a continually-evolving process as the technology itself develops, but we are acting now by accepting in full all 44 recommendations of the Independent Commission, which will provide a foundation to embrace the opportunities AI brings while putting strong safeguards, transparency and public confidence at the heart of its use.'
As part of this process, the MHRA is inviting applications for the third phase of AI Airlock Phase 3 which working with developers, regulators and healthcare partners to test how AI devices can be safely monitored and managed for post-market surveillance after they are deployed . Applications are now open for developers of eligible AI-enabled medical devices, with a webinar for prospective applicants taking place on 22 October.
The MHRA also committed to issue draft guidance by December 2026 on a new approach to managing changes to AI-enabled medical devices as they adapt and improve over time and will also begin next year by consulting on how to qualify and classify AI-enabled devices, the first recommendation of the National Commission.
Other commitments include exploring 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.
Reaction
Stuart Harvey, chief executive of Datactics, commented: ‘Constant testing is what keeps AI in check and models must be tested on appropriate data, in a controlled setting with clear and enforced limits, and then retested regularly to ensure they are behaving and performing as intended. A safe space is only as safe as the data inside it and if an AI system is built on inaccurate data, or operates outside its prescribed parameters, even the smallest of error can have devastating impact on patient care.'
Dr Hannah Allen, chief medical officer at Heidi, said: ‘Greater transparency and predictability for patients are central to the recommendations, and that's what will build patient trust as these tools are used more widely. What matters now is getting implementation right, so that technology like ambient voice can reduce the amount of time healthcare professionals spend on admin so they can spend more time with patients.'
Gerard Hanratty, partner and head of health and life sciences at UK and Ireland law firm Browne Jacobson, said: ‘The proposed L-plates-style staged authorisation model is a pragmatic and proportionate response to the genuine challenge of deploying adaptive technologies safely in real-world NHS settings. Equally important is the recommendation for continuous, lifecycle monitoring of AI-enabled medical devices, which reflects the reality that these are not static products. They evolve after deployment, and our regulatory frameworks must keep pace with that.'
