Summary:
Le programme AI Airlock de la MHRA a ouvert ses candidatures pour une deuxième phase, avec un financement de 1 million de livres pour tester des dispositifs médicaux basés sur l’IA. Cette initiative permet aux entreprises de collaborer avec les régulateurs dans un environnement contrôlé, facilitant l’évaluation et l’amélioration des technologies de santé avant leur déploiement. Le programme vise à équilibrer la réglementation stricte tout en favorisant l’innovation, renforçant ainsi la position du Royaume-Uni en tant que leader mondial de l’innovation en santé.
Original Link:
Original Article:
The MHRA has opened applications for the second round of its AI Airlock programme, backed by £1m to expand regulatory testing of AI-powered medical devices.
The scheme allows companies to work with regulators in a controlled ‘sandbox’ environment, where AI-driven health technologies can be safely tested and refined before wider rollout. Applications opened on 23 June 2025 and will close on 14 July 2025.
The second round follows a pilot phase that tested four AI technologies, including software to help clinicians create personalised cancer treatment plans, and a tool to monitor AI performance in real time in clinical settings.
Launched by the Medicines and Healthcare products Regulatory Agency (MHRA), the programme is described as a first-of-its-kind regulatory testing space. Like an airlock in spaceflight, the sandbox creates a boundary between experimental AI tools and fully approved medical devices.
The initiative builds on commitments in the government’s AI Opportunities Action Plan and its response to the Regulatory Horizons Council’s report on AI regulation. It is supported by the Regulatory Innovation Office (RIO), which is helping regulators adopt more flexible and responsive approaches to emerging technologies.
James Pound, MHRA interim executive director for innovation and compliance, said: “Traditional regulatory pathways weren’t designed with AI’s unique characteristics in mind – including its capacity to analyse large quantities of data and help automate existing manual processes. The AI Airlock programme helps address this gap by creating a supervised testing ground where these novel technologies and challenge areas can be safely investigated.
“The technologies and devices which have been evaluated to date have shown the limitless potential of AI to improve patient outcomes, free up NHS resources, and enhance the accuracy and efficiency of healthcare services.
“With AI, we must balance robust oversight with flexibility that doesn’t stifle innovation, and this programme achieves that balance.”
Science minister Lord Vallance said: “Backing innovation means backing better regulation – and that’s what the RIO is here to do.
“Smarter, faster approaches like the AI Airlock are helping to cut red tape, bring safe new technologies to patients quicker, and ease pressure on our NHS – fuelling the government’s Plan for Change.”
Health minister Baroness Merron added: “AI has huge potential to improve healthcare, and we need to use it safely and responsibly. The AI Airlock programme is a great example of how we can test new technology thoroughly while still moving quickly.
“This £1m investment will help bring new medical tools to patients faster and strengthen the UK’s position as a global leader in healthcare innovation.”
Among the first cohort was Philips’ Radiology Auto Impression project, which tested generative AI to automate the ‘impressions’ section of radiology reports – a critical summary of key findings from imaging procedures.
Yinnon Dolev, Philips’ advanced development NLP tech lead, said the weekly meetings with MHRA specialists were “almost unheard of” and provided “a catalyst for meaningful progress expediting our development activities.”
Another project, OncoFlow, developed AI to support healthcare professionals in creating personalised cancer patient management plans. The technology aims to reduce delays to cancer appointments, potentially enabling earlier treatment and improving survival outcomes.
Co-founder Aruni Ghose said the programme allowed his team to validate their product in a simulated clinical environment and “pressure-test it against real regulatory standards,” helping the company progress “from idea to a validated MVP (minimum viable product).”
Two further pilot participants included Automedica Ltd, which examined the regulatory use of retrieval-augmented generation (RAG) – a method combining large language models with verified data sources – and Newton’s Tree, which tested its Federated AI Monitoring Service (FAMOS) to detect AI-related risks such as performance drift or safety issues in clinical settings.
Findings from all four projects are expected later this year and will help shape future iterations of the Airlock programme, as well as broader regulatory approaches to AI in healthcare.
To qualify for the second cohort, applicants must show that their AI-powered medical device could deliver significant benefits to patients and the NHS, present a novel treatment approach, and pose a regulatory challenge appropriate for sandbox testing.
The AI Airlock was launched in spring 2024 and is the MHRA’s first regulatory sandbox focused on AI as a medical device (AIaMD) products.