UK MHRA News: "AI Airlock" Pioneering Safe Healthcare Innovation in the UK

The UK’s Medicines & Healthcare products Regulatory Agency (MHRA) has launched Phase 2 of its AI Airlock programme, supported by a £1 million investment. This initiative marks a ground‑breaking step in accelerating safe, effective AI‑powered medical tools for patient care.

What Is the AI Airlock Programme?

The AI Airlock is the UK’s first regulatory sandbox for AI as a Medical Device (AIaMD). Much like an airlock on a spacecraft, this “sandbox” creates a safe boundary between experimental AI technologies and fully approved medical devices used in real-world healthcare.

• Regulatory sandbox: A structured, controlled environment allowing innovators to test AI as a Medical Device (AIaMD) under real‑world conditions while regulators guide compliance and risk mitigation.

• Pilot success: Phase 1 proved the concept’s worth by identifying regulatory challenges early, enabling quicker, safer product adaptations.

• Phase 2 expansion: The MHRA is now selecting a new cohort, including projects like Philips’ generative‑AI tool for generating radiology report impressions, highlighting the practical impact of AI in diagnostics

• Companies selected for the programme work directly with MHRA regulators to test their innovations, identify regulatory challenges early, and refine their products for safe, effective use in clinical settings.

  
  

Why Is This Programme Important?

Traditional regulatory frameworks were not designed for the unique challenges posed by AI technologies, which can rapidly analyze vast data sets and automate complex processes. The AI Airlock addresses this gap by offering:

• A supervised, real-world testing environment

• Early identification and resolution of regulatory hurdles

• Opportunities for companies to collaborate with regulators and healthcare professionals

• A pathway to faster, safer patient access to innovative tools1

This approach supports the UK Government’s broader AI Opportunities Action Plan and aligns with the Regulatory Horizons Council’s recommendations for agile regulation in the AI era.

Highlights from the Pilot Phase

The inaugural AI Airlock cohort featured four breakthrough projects, each tackling critical healthcare challenges:

• Philips’ Radiology Auto Impression Project: Used generative AI to automate the writing of radiologists’ final impressions—summaries of key findings in imaging reports. The project benefited from direct, weekly collaboration with MHRA experts, helping define robust testing strategies and accelerating development.

• OncoFlow: Developed AI tools to create personalized cancer management plans, potentially reducing waiting times for cancer appointments and improving survival rates. The Airlock enabled OncoFlow to validate its product in a simulated clinical setting and meet real regulatory standards.

• Automedica Ltd: Explored the regulatory advantages of retrieval-augmented generation (RAG) technologies, combining verified knowledge bases with large language models (LLMs).

• Newton’s Tree’s Federated AI Monitoring Service (FAMOS): Tested AI systems for identifying and mitigating risks in clinical settings, such as performance drift or safety issues.

The results from these pilot projects will be published later this year, with insights expected to shape future regulatory approaches for AI in healthcare.

For more details, you can read the full announcement on the UK Government’s official website here.