The UK’s acceptance of 44 healthcare AI recommendations is a policy commitment, not evidence that every proposed reform has taken effect. The government accepted the National Commission into the Regulation of AI in Healthcare’s recommendations on 6 October 2026 and announced the next phase of AI Airlock, the Medicines and Healthcare products Regulatory Agency’s (MHRA) sandbox. Government announcement
For healthcare providers and device developers, the immediate distinction is between a policy commitment, guidance still to be drafted and a testing programme already accepting applications.
What the response commits to
The government’s action plan includes draft guidance on predetermined change control plans: a way to define the boundaries within which a device could change while retaining its intended use. That draft is expected by December 2026.
The MHRA also plans to consult on qualification and classification of AI-enabled devices. Separately, it will explore staged authorisation pathways, potentially allowing evidence to develop alongside closely supervised use. The response describes options to develop; it does not establish that a new staged pathway is available to manufacturers now.
The announcement has been corrected to refer to ten priority areas, rather than nine. A fuller implementation roadmap is due by spring 2027. Delivery announcement and correction
What AI Airlock is testing now
The Phase 3 application guidance focuses on monitoring devices after deployment and across their working life. Its questions include how local workflows and data quality affect performance, how monitoring should work where clinical oversight is limited, and how updates or rollbacks reach the people who need to know.
Applications remain open on a rolling basis. The current review wave asks applicants to apply before 31 October; the window continues afterwards. Participation has no fee, but applicants are expected to fund their own studies and testing work.
The programme’s real-world testing environment observes systems outside patient pathways, such as in shadow deployment. Outputs are not used to make actual patient decisions in that environment. Sandbox participation should therefore not be read as a product authorisation.
The next evidence to watch
The next concrete documents are the change-control draft, the classification consultation and the implementation roadmap. Each will show more precisely how the government intends to turn its accepted recommendations into practice. The UK commission response and AI Airlock call provide no basis for claiming that this package has already improved clinical outcomes.