A £1 million boost to the Medicine and Healthcare Products Regulation Authority (MHRA) pioneering AI Airlock Programme will allow businesses to work directly with regulators to explore ways to safely test new AI-powered medical devices and improve patients faster and regular adjustments.
The second round of programs open today (June 23, 2025) application follows a successful pilot phase where it saw four groundbreaking AI technologies, including software that helps doctors create personalized cancer treatment plans, and software that helps hospitals, AI developers and regulators monitor AI performance in real time.
Like spacecraft airlocks, the “sandbox” test space creates the boundaries between experimental AI and fully approved medical technology used in the real world.
The initiative is based on the government's commitment to the AI Opportunity Action Plan and its response to the Regulatory Horizons Council on regulating AI as a medical device to enable safe AI innovation through strategic guidance to regulators and enhance AI capabilities.
The program is supported by the government's new Office of Regulatory Innovation (RIO). It supports regulators to test more agile and flexible ways of working that can keep up with new technologies like AI. By reducing unnecessary deficits and making the UK a more innovation-friendly environment, Rio will help realize its government's plan for change – supporting high-growth industries, supporting NHS innovation, and promoting technology that will make a real difference in people's lives.
The Minister of Science, Balance Lord said:
“Back innovation means supporting better regulations – that's what Rio is here.
“A smarter, faster approach like AI airlocks will help reduce red tape, bring new technology that is safer for patients more quickly, and relieve pressure on the NHS.
Health Minister Barones Mellon said:
“AI has great potential to improve healthcare and needs to be used safely and responsibly. The AI airlock program is a great example of how to thoroughly test new technologies while moving quickly.
“This £1 million investment will help bring new medical tools to patients more quickly and strengthen the UK's position as a global leader in healthcare innovation.”
Selected for the next round of the AI Airlock program can test AI healthcare products under careful supervision, identify and adjust regulatory challenges early.
“We're excited to announce that we're in the process of innovation and compliance,” said James Pound, Interim Executive Director of MHRA.
“Traditional regulatory pathways were not designed with the unique properties of AI in mind that help analyse large amounts of data and automate existing manual processes. The AIAIRLOCK program will help address this gap by creating a monitored test site that can safely investigate these new technologies and challenge areas.
“Technology and devices that have been evaluated up to date demonstrate the limitless possibilities of AI, improving patient outcomes, freeing up NHS resources, and increasing the accuracy and efficiency of health services.
“With AI, we need to balance robust surveillance and flexibility that doesn't stifle innovation, and this program achieves that balance.”
Four projects were selected for the first AI airlock cohort, each focusing on addressing critical healthcare challenges using AI. Among them is the automatic impressions project of radiology by health technology multinational Philips, which tested the use of generated AI to automate the writing of radiologists' final impressions.
Working directly with MHRA experts through weekly meetings, the team gained valuable insight into the need to involve end users (radiologists) to help define testing strategies. As Inon Dref, lead Phillips' advanced development NLP technology, collaboration with regulators was “almost unprecedented” and provided “a catalyst for meaningful advancement that promotes development activities.”
Another first round project, Oncoflow, looked at the use of AI to help healthcare professionals develop personalized management plans for cancer patients, which could reduce wait times for cancer appointments, and significantly increased early treatment and patient survival. Co-founder Aruni Ghose said the Airlock program will help his team validate products in a simulated clinical setting, provide “pressure tests against real regulatory standards” and drive progress from IDEA to validated MVPs (minimum viable products).
Finishing the cohort was two projects. One, according to Automedica Ltd, will investigate the regulatory benefits of using searched generation (RAG) technologies using a searched knowledge base and a large-scale language model (LLM). The other is testing the federated AI monitoring service (FAMOS) by health technology startup Newton Tree, identifying and mitigating AI risks in clinical settings that include performance drift and safety issues.
The results of all four pilot projects will be released later this year, providing valuable insights to help advance the AI airlock program and inform the broader regulatory approach to effective and safe use AI in healthcare.
Eligible candidates in the second cohort should demonstrate that AI-powered medical devices can provide significant benefits to patients and the NHS, presenting new treatment approaches and providing regulatory challenges that are ready to be tested in airlock programs.
Applications for Cohort Two Open on June 23, 2025 will end on July 14, 2025.
