MHRA's AI Airlock Points to Lifecycle Regulation for Medical AI
The UK regulator's latest AI Airlock phase argues that medical AI oversight cannot stop at pre-market validation, especially when model behaviour, human oversight and clinical context can change after deployment.
Illustrative image. Cedar S. Insights uses editorial stock photography; images do not depict specific events described in articles.
The Medicines and Healthcare products Regulatory Agency has published insights from Phase 2 of its AI Airlock programme, a regulatory sandbox for AI medical devices.
The programme ran from April 2025 to March 2026 and worked with seven AI technologies selected from 51 applicants. The cohort covered clinical note-taking and summarisation, advanced cancer diagnostics, rare eye disease detection and obesity management support systems.
The MHRA says the work focused on three regulatory challenges: managing intended purpose and validation for generative AI medical devices, lifecycle management through predetermined change control plans and post-market surveillance, and performance evaluation for AI-powered in vitro diagnostics.
Several findings matter beyond the UK. The agency says pre-market validation alone may not be enough because real-world performance can be difficult to reproduce in controlled settings. It also warns that human oversight changes over time: as users grow more confident in a system, they may scrutinise outputs differently. For generative AI and large-language-model systems, behaviour may shift even without explicit design changes, creating risks if guardrails and intended-use boundaries are weak.
The MHRA argues that clinical relevance should underpin performance metrics, because statistical significance does not automatically mean clinical importance. It also calls for a broader ecosystem view, since not every healthcare AI tool falls neatly within the medical device definition even when it affects patient care.
The AI Airlock is a sandbox, not a binding regulatory framework. The insights published represent the agency's current thinking from working with a small cohort of technologies. They should not be read as final regulation.
Why It Matters
This is the governance side of AI adoption. As NHS and health systems scale AI tools, regulators are moving toward continuous oversight: performance monitoring, change control, human factors and clinically meaningful thresholds. That is slower than hype, but much closer to how high-stakes healthcare technology actually has to work.
Our sourcing: Cedar S. Insights provides source-led editorial analysis. Reported company, institutional and regulatory claims are attributed to their original sources unless stated otherwise.
Corrections: If a material factual error is identified, Cedar S. Insights will update the relevant article and preserve the distinction between the corrected statement and supporting evidence.
Topics