AI Airlock: the regulatory sandbox for AIaMD
A proactive, collaborative, agile and the first of its kind approach to identifying and addressing the challenges faced by AI as a Medical Device (AIaMD).
Why we are setting up AI Airlock
As the medical device regulator for UK health and social care, the Medicines and Healthcare products Regulatory Agency (MHRA) holds a central role in patient safety. We are currently reforming the medical device regulations and supporting regulatory guidance, including for Software as a Medical Device (SaMD) which includes Artificial Intelligence as a Medical Device (AIaMD) products. These reforms build on a well-established regulatory system of legislation, guidance and standards to address known challenges and align with other regulatory jurisdictions around the world.
Along with this regulatory system update, we have seen a marked increase in innovative devices entering the UK market, particularly AIaMD and digital technologies using AI, which may be on the borderline of the scope of medical device regulations. Such products have the potential to revolutionise the healthcare landscape, maintain the UK as a home of cutting-edge research and provide improved ways to address patient and public needs. This potential brings novel challenges, shifts in known risks and safety concerns. The MHRA is seeking to balance appropriate oversight to protect patient safety with the agility needed to respond to the particular challenges presented by these products to ensure regulation does not present undue barriers to innovation.
Many of the known risks of SaMD/AIaMD products are identified and mitigated through existing regulatory requirements and processes such as Clinical Investigations and Post Market Surveillance. Our existing publications and further planned guidance documents outlined in the Software and AI as a Medical Device Change Programme provide further assistance to manufacturers. The pace of innovation is accelerating and the balance of rules-based versus principles-based regulation of AI is an ongoing topic across all sectors of society and around the world. To regulate AIaMD effectively and efficiently, increased collaboration and more proactive and agile ways of working are required, to help us find solutions to many unanswered questions.
This webpage contains information about the project, the key partners and some examples of potential issues we expect to address in the project. We will be holding a webinar for innovators and developers on 5 June 2024 to provide further information and answer questions relating to the project.
Explaining the AI Airlock
The MHRA AI Airlock project is now underway and we will launch a call for product applications shortly. The project will enhance our understanding and accelerate solutions to novel regulatory challenges for AIaMD.
As part of the project scoping and setup, we have reviewed the achievements of other regulators through the use of regulatory sandboxes and we intend to build on lessons learned by others. A core difference between the AI Airlock and other regulatory sandboxes is the need for collaboration across many of the regulatory, governance and assurance organisations in healthcare. Using real-world products, the AI Airlock will bring together expertise from within the MHRA and key partners including the UK Approved Bodies, the NHS and other regulators. The outputs will inform subsequent AI Airlock phases in the short term, and future MHRA guidance and policy in the longer term, while exploring any limitations of existing approaches to demonstrating regulatory compliance.
Collaborating with partners
Team AB: launched in February 2024, Team AB brings together expertise from Approved Bodies to increase consistency of interpretation of UK medical device regulatory requirements. We are delighted to collaborate with Team AB on the AI Airlock project to uncover and navigate medical device regulatory challenges for AIaMD. We will use our collective regulatory knowledge to inform and standardise policy positions in this rapidly evolving area.
The NHS: the AI Airlock project will work closely with the Department of Health and Social Care (DHSC), and the NHS AI Lab, to provide connections and expertise with our healthcare system. Many AIaMD products are deployed via NHS infrastructure making NHS England and the health services in the devolved nations crucial to regulatory discussions such as deployment and post market surveillance.
Project outputs
The objective of AI Airlock is to identify regulatory challenges to AIaMD and work collaboratively to understand and potentially mitigate any risks that are uncovered. Relevant stakeholders and the product providers will be able to use the AI Airlock product reports to share knowledge and findings to assist with further funding or assessment activities. Whilst guarantee of regulatory conformity is not the objective of this project, involvement in this project will benefit manufacturers on their regulatory journey. The MHRA also intends to engage academic expertise to assess the multi-partner approach to the use of regulatory sandboxes and its impact on the innovation landscape. Finally, extension and additional funding, potential changes to medical device regulations and additional guidance documents will be considered for further sandboxes looking beyond AIaMD.
How to get involved in AI Airlock
The pilot project will focus on a small number of products, across a range of medical device regulatory issues, with different sections of healthcare or clinical disciplines and levels of product regulatory maturity. Some such regulatory challenges could include, amongst others, understanding the safety, validation and design implications of:
- Detecting and reporting product performance errors (including drift) and failure modes in post market surveillance data.
- Increased automation and decision-making responsibilities within clinical workflow and producing pre-market evidence of safety.
- Breaking down the complexities of generative AI based medical devices.
AI Airlock will open for applications after the project webinar on 5 June 2024. Applications will be sifted and prioritised against the application criteria and 4-6 will be selected for the pilot project. Each product airlock will be set up and tailored to the specific needs of the innovation, to be flexible to potential issues and to maximise opportunities to explore concerns within acceptable risk levels. Products may start in one category of the AI Airlock and progress to others, or if a sufficient safety concern is identified, removed from the activity altogether. Successful applicants to the AI Airlock will be required to work collaboratively and transparently with the project team. Afterwards, all parties will produce reports from their perspectives, summarising the achievements and lessons learnt to inform regulatory policy and guidance.