The UK has opened applications for a new program to test how AI medical devices can be monitored after they enter use. The Medicines and Healthcare products Regulatory Agency (MHRA), which regulates medicines and medical devices, announced the third phase of its AI Airlock sandbox on October 6, the same day the government accepted all 44 recommendations from its healthcare AI commission.
AI Airlock Phase 3 will focus on post-market surveillance: checking a device’s safety and performance after deployment, as well as how it behaves in the settings where clinicians use it. The MHRA wants participating developers, regulators and healthcare partners to test ways of spotting problems that a pre-deployment assessment may not reveal, such as effects from local workflows, infrastructure, data quality or user interaction.
The commission’s September report called for oversight across a product’s life, rather than relying mainly on a one-time check before use. That matters for AI tools that can be updated or whose results depend on the data and conditions at each hospital. The government response turns that recommendation into a set of priorities, but it does not itself create new medical-device rules.
The response names ten areas for work, including modernising how AI-enabled devices are classified, strengthening monitoring after deployment and making safe product changes easier to manage. The MHRA plans to publish draft guidance on change-control plans by December 2026 and consult on qualification and classification by Spring 2027. A cross-system board will meet quarterly; a fuller roadmap with responsibilities and timelines is also due by Spring 2027. The government says it is exploring supervised, staged routes that could let some promising tools reach NHS use sooner while more evidence is collected. That pathway has not yet been launched.
Eligible developers can apply to AI Airlock throughout the year. Applications for the current review wave are due by October 31, with the first participants expected to be selected in November. The phase is funded through April 2029, and individual testing engagements are expected to last six to 12 months. One option, the programme’s “real-world airlock,” assesses a device in its intended clinical setting using retrospective patient data. It is a shadow deployment, so the device’s output does not inform real treatment decisions.
The immediate change is a place to test monitoring approaches, not proof that any device is safe or effective, and not a new approval for products already on the market. The government has accepted the commission’s recommendations; many details still depend on future guidance, consultation and the Spring 2027 roadmap.
Image: GOV.UK / MHRA. Reused under the Open Government Licence v3.0.