Earning Trust for AI in Health 2025

Exceptions to market access practices in healthcare Regulatory sandboxes in healthBOX 1 BOX 2Health regulators already adapt their regulatory practices in emergency situations. For instance, many regulators grant access to drugs before market authorization for patients suffering from a serious illness who have no viable treatment options available to them: –Expanded Access Program (EAP) in the United States –Temporary Authorization for Use in France –Early Access Medicine Scheme in the United Kingdom –Exceptional Use Authorization (EUA) for medical devices in the United KingdomThese schemes operate under regulatory oversight and are often time-limited and evidence-dependent. Similarly, dedicated frameworks support provisional access to digital health technologies while evidence is still being generated: –PECAN (France): pilot allowing provisional access for digital tools with ongoing data collection –DiGA (Germany): provisional fast-track reimbursement for digital health apps –Early Value Assessment (United Kingdom): rapid assessment of technologies addressing unmet needs A sandbox is a framework created by health authorities or regulatory agencies to allow healthcare innovators to test new digital health technologies with tailored regulatory constraints. A sandbox often provides resources, such as datasets or advisory support, to assist innovators and can even extend to creating digital public goods – open access software or data intended to contribute to sustainable digital development, for example.Sandboxes for AI can: –Enhance the understanding of AI solutions before they enter the market –Support the development of effective enforcement policies and technical guidance to mitigate risks –Foster AI innovation by providing a controlled testing environment for emerging AI technologies26Dedicated testing environments such as regulatory sandboxes can help promote the development of high-quality AI technologies in health (see Box 2 for definition). Sandboxes must be adapted to the context and environment in which they operate. In countries with comprehensive regulatory frameworks, sandboxes can focus on tailoring or modifying regulatory provisions and processes, while they should provide a foundational framework to support innovation in countries (Lebanon and Pakistan, for example) that are still developing a clear regulatory framework.24 Furthermore, it is important that regulatory sandboxes are focused on a specific industry (e.g. healthcare) so that any sector-specific requirements can be embedded in the design of the sandbox (such as in Portugal).25 Carefully designed regulatory sandboxes can help innovators collect insights on the real-world performance of AI technologies in health while ensuring that patient safety and privacy are protected within a clearly defined environment. 2.2 Sandboxes provide a safe space in which the private sector can innovate For example, the United Kingdom’s Medicines and Healthcare products Regulatory Agency (MHRA) launched AI Airlock in spring 2024, its first regulatory sandbox for AI as a Medical Device (AIaMD). The goal of the project is to understand and accelerate “solutions to novel regulatory challenges for AIaMD” due to a marked increase in innovative devices entering the UK market. 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 that regulation does not present undue barriers to innovation”.27 Earning Trust for AI in Health: A Collaborative Path Forward 11