Summary

  • The UK government accepted all 44 recommendations on AI in healthcare, but the working group charged with clarifying responsibilities is expressly barred from redistributing them.
  • A Spring 2027 implementation roadmap must turn the cross-system pledge into roles, dates and deliverables; MHRA’s new AI Airlock phase offers an operational test of post-market monitoring.

Acceptance starts the delivery phase

The UK’s 6 October response to the National Commission into the Regulation of AI in Healthcare accepts all 44 recommendations. It sets priorities for modernising medical-device rules, improving system readiness and shifting towards stronger post-market assurance. The response says current UK Medical Device Regulations 2002 are no longer fit for purpose for software and AI products as they are developing, and MHRA intends to consult by Spring 2027 on a revised device definition, risk-based classification and how design, function and context affect a product’s intended purpose.

That is a programme of work, not a new device law. The response also promises a cross-system Programme Board that will meet quarterly. A delivery plan and roadmap is due by Spring 2027, with more detail on roles, responsibilities, timelines and key deliverables, followed by annual progress updates.

The group can clarify a boundary, not move it

Recommendation 24 addresses who does what when an AI product crosses the line from a manufacturer’s system into a health service. The Department of Health and Social Care, MHRA, Care Quality Commission, devolved counterparts, NHS Resolution, the Professional Standards Authority and professional regulators are to form a working group to make responsibilities clearer across manufacturers, providers, professionals and regulators.

The response sets a limit: the group “will not determine or redistribute responsibilities.” That leaves it to work within existing statutory remits. The government says DHSC and devolved health departments are separately considering governance practices for AI tools, including those with increasing degrees of autonomy. It gives no detailed allocation or timeline for that work beyond further details by Spring 2027.

This is not evidence that no one currently has a duty. It is a warning that more AI-specific coordination is needed across duties held by different bodies. The distinction matters when a device is technically compliant but performs differently after a hospital changes its workflow, a provider’s data quality shifts or a product is used with less clinical oversight.

Airlock moves the test into deployment

On the same day as the response, MHRA opened applications for Phase 3 of its AI Airlock sandbox. The current wave focuses on post-market surveillance and oversight over a device’s lifecycle. Its application asks whether manufacturers can capture local infrastructure, workflow integration, data availability and quality, and user interaction; detect safety or performance signals where oversight varies; and communicate performance information, safety issues, updates and rollbacks to regulators, providers, professionals and patients.

The call is open year-round. Applicants seeking the current review wave are asked to apply by 31 October; another shortlisting period is planned for Spring 2027. These are programme questions, not findings: no Phase 3 cohort or result has yet been published.

The gap is therefore concrete. A post-market signal is useful only if someone can access the relevant data, judge whether a change matters and trigger an agreed response. MHRA’s sandbox is designed to explore those mechanics; the government’s cross-system work must connect them to the people who can act.

Sources