Medical Device Regulatory Sandboxes: MHRA AI Airlock, FDA TEMPO, and How Innovation Programs Are Reshaping Device Approvals in 2026

The Regulatory Sandbox Concept Comes to Medical Devices

Traditional medical device regulation was built for static products — a pacemaker, a surgical instrument, an imaging system — that do not fundamentally change after market entry. But AI-enabled diagnostics, adaptive software, and connected health devices evolve continuously. They learn from new data, update their algorithms, and expand their capabilities over time. The existing regulatory framework struggles to keep pace with this evolution, creating a growing gap between what technology can do and what regulation allows.

Regulatory sandboxes are the response. Borrowed from the financial services sector — where the UK's Financial Conduct Authority pioneered the concept in 2016 — sandboxes create controlled environments where manufacturers can test innovative products with real patients under regulatory supervision, without first obtaining full market authorization. Regulators learn alongside manufacturers, developing evidence-based frameworks for technologies that do not fit neatly into existing rules.

In 2025–2026, three major sandbox programs have emerged that medical device manufacturers need to understand: the UK MHRA's AI Airlock, the US FDA's TEMPO pilot, and the EU's proposed sandbox provisions within the MDR/IVDR targeted revisions. Together, they represent a fundamental shift in how regulators approach innovation — from gatekeeping to collaboration.

MHRA AI Airlock: The World's First AI Medical Device Sandbox

Background and Launch

The MHRA launched the AI Airlock in 2024 as the world's first regulatory sandbox specifically designed for Artificial Intelligence as a Medical Device (AIaMD). The program brings together regulators, the NHS, industry partners, and the UK Association of Medical Device Approved Bodies (Team-AB) to explore how AI-enabled medical devices can be evaluated, approved, and monitored when the technology itself is designed to evolve.

Phase 1 (Pilot, 2024)

The pilot phase tested initial concepts and identified key regulatory challenges unique to AI medical devices:

• Risk of errors and inaccuracies from AI models generating outputs not grounded in validated clinical data
• Transparency of recommendations — ensuring clinicians understand why an AI system makes a specific recommendation and can maintain confidence in the tool
• Post-market monitoring — detecting changes in AI performance after deployment, including model drift, data distribution shifts, and over-reliance by users

The pilot produced reports and case studies published on October 16, 2025, establishing foundational knowledge for the program's expansion.

Phase 2 (2025–2026)

Phase 2 expanded dramatically with seven additional AI-enabled technologies, selected to address three core regulatory challenges:

Challenge 1: Adaptive, evolving, and self-learning AI-enabled digital health technologies

• How do you regulate a device whose algorithm continuously learns and changes?
• What constitutes a "significant change" when the device is designed to evolve?

Challenge 2: AI-powered diagnostics

• How should performance metrics for AI-driven diagnostic tools be defined and measured?
• How can performance be aligned with international standards while remaining adaptable?

Challenge 3: Post-market surveillance for AI devices

• How should ongoing monitoring work for devices whose behavior may change over time?
• How can predetermined change control plans (PCCPs) be integrated into post-market frameworks?

Phase 2 candidates span diverse technologies: AI-powered clinical note-taking (using large language models), advanced cancer diagnostics, eye disease detection, and obesity treatment support systems. Candidates test in three environments:

Phase 2 testing runs through April 2026, with reporting expected Summer 2026.

£3.6 Million Multi-Year Funding (2026–2029)

In April 2026, the Department of Health and Social Care allocated £1.2 million per year for three years (2026–2029) to the AI Airlock. This transition from annual to multi-year funding signals that the UK government sees AI device regulation as a sustained strategic priority, not a pilot experiment.

Key implications of the funding expansion:

• More technologies can enter the sandbox in each cohort
• Longer testing periods enable more thorough evaluation of adaptive AI systems
• International collaboration becomes feasible, with the MHRA positioned as a global leader in AI regulatory innovation
• Phase 3 design will be informed by Phase 1 and Phase 2 findings, shaping the MHRA's broader regulatory approach to AI

James Pound, MHRA Executive Director of Innovation and Compliance, stated: "Securing this multi-year funding boost marks a pivotal moment for AI Airlock and for the safe and responsible advancement of AI in healthcare."

Phase 2 Candidates

(Note: Full candidate list includes 7 technologies; some details remain confidential during active testing.)

FDA TEMPO Pilot: A New Model for Digital Health Access

What Is TEMPO?

The Technology-Enabled Meaningful Patient Outcomes (TEMPO) pilot, announced December 5, 2025, and opened for applications January 2, 2026, represents the FDA's most ambitious sandbox-style program to date. Developed by the Center for Devices and Radiological Health (CDRH) in collaboration with the Centers for Medicare and Medicaid Services (CMS) Innovation Center, TEMPO links regulatory access with payment reform — a combination no other regulator has attempted.

How TEMPO Works

TEMPO allows manufacturers of certain digital health devices to request FDA enforcement discretion for premarket authorization requirements, Investigational Device Exemption (IDE) requirements, and certain informed consent provisions (21 CFR Parts 50 and 56). In exchange, manufacturers collect, monitor, and report real-world performance data.

The program is tied to CMS's Advancing Chronic Care with Effective, Scalable Solutions (ACCESS) model, a new payment initiative launching July 5, 2026 that supports technology-enabled chronic care for Medicare beneficiaries with outcome-aligned payments.

Four Clinical Use Areas

TEMPO focuses on four chronic disease categories:

Eligibility Criteria

Devices eligible for TEMPO must:

• Be digital health devices intended for use in clinician-supervised outpatient settings
• Be designed to improve patient outcomes in one of the four clinical use areas
• Be offered for an intended use that aligns with care covered by the CMMI ACCESS model
• Not be a device that requires premarket approval (PMA) or 510(k) clearance (enforcement discretion applies)

Key Timeline

How to Apply

Manufacturers submit a statement of interest to FDA-TEMPOPilot@fda.hhs.gov titled "Statement of Interest for Participation in the TEMPO Pilot." The statement should include:

1. Manufacturer identification and contact information
2. Device description and proposed indications for use
3. Request for FDA to exercise enforcement discretion
4. Explanation of how the device aligns with one of the four clinical use areas
5. Plan for collecting, monitoring, and reporting real-world performance data

The FDA plans to select up to approximately 10 manufacturers in each of the four clinical use areas. As of April 2026, the FDA has begun sending follow-up requests to selected potential participants and continues to accept statements of interest.

Why TEMPO Matters

TEMPO is the most innovative sandbox program globally because it simultaneously addresses two barriers that have historically slowed digital health adoption:

1. Regulatory barriers — Manufacturers can deploy devices under enforcement discretion rather than waiting for full premarket review
2. Payment barriers — The ACCESS model provides outcome-aligned payments, so manufacturers and clinicians have financial incentives to use the devices

FDA Commissioner Marty Makary described the program: "We are piloting an approach to encourage the use of digital technologies that meet people where they are. This pilot supports innovative tools and a health care delivery model that could improve care for millions of Americans managing chronic disease."

EU Sandbox Proposals: MDR/IVDR Targeted Revisions

The Commission's Sandbox Vision

The European Commission's December 16, 2025 proposal for targeted revisions to the MDR and IVDR (COM(2025) 1023) includes explicit provisions for regulatory sandboxes. While less developed than the MHRA or FDA programs, the EU approach would create a framework for member states to establish sandboxes for innovative medical devices — particularly AI-enabled technologies.

Key Proposed Provisions

The revisions envision sandboxes that would:

• Allow manufacturers to test innovative devices in controlled settings with regulatory oversight
• Enable competent authorities to develop expertise in emerging technologies
• Create evidence bases for future regulatory frameworks
• Support the development of harmonized standards for novel device categories

The proposal is currently in the consultation phase (feedback open until May 6, 2026), with adoption expected by the European Parliament and Council around Q2 2027.

Current EU Innovation Support

While formal sandbox provisions are still being negotiated, the EU has existing mechanisms that serve similar functions:

• EU Innovation Framework — Early dialogue with expert panels for novel devices
• Clinical Investigation Consultation — Pre-submission engagement with competent authorities
• Notified Body Innovation Offices — Several NBs have established innovation teams to advise manufacturers on novel technologies
• Horizon Europe funding — Supports development of innovative medical technologies

Other Global Sandbox and Innovation Programs

Singapore HSA

Singapore's Health Sciences Authority (HSA) has operated informal sandbox arrangements for novel digital health technologies, particularly through its regulatory guidelines for AI medical devices. The approach is pragmatic: HSA works directly with manufacturers on a case-by-case basis to define appropriate evidence requirements.

South Korea MFDS

South Korea's "Market Immediate Entry Medical Technology" fast-track, launched January 26, 2026, shares sandbox principles. It reduces approval timelines from 490 days to 80–140 days for innovative devices across 199 categories, with enhanced post-market monitoring requirements.

Australia TGA

The Therapeutic Goods Administration has indicated interest in sandbox-style approaches for AI medical devices, though no formal program has been launched as of May 2026.

How Sandbox Participation Benefits Manufacturers

For AI and Digital Health Companies

For Traditional Device Manufacturers

Even manufacturers of non-AI devices should pay attention to sandbox programs:

• Sandbox findings shape future regulatory expectations for all devices
• Programs often develop new approaches to clinical evidence, post-market surveillance, and quality management that eventually become standard requirements
• The EU MDR/IVDR revisions are directly informed by sandbox learnings from MHRA and FDA programs

Practical Considerations for Manufacturers

Should You Apply?

Consider sandbox participation if your device:

• Involves AI, machine learning, or adaptive algorithms
• Targets chronic disease management in one of TEMPO's four clinical areas (for FDA)
• Involves novel technology that doesn't fit neatly into existing classification rules
• Could benefit from real-world evidence collection in controlled settings
• Is at a stage where regulatory engagement would de-risk future submissions

What Participation Requires

• Resource commitment: While the MHRA AI Airlock has no participation fee, candidates fund their own studies and testing
• Transparency: Sandbox programs typically require publication of findings
• Time: Testing periods range from 6 months (MHRA) to ongoing (TEMPO)
• Cross-functional engagement: Regulatory, clinical, engineering, and data science teams must collaborate

Common Misconceptions

Myth: Sandbox participation replaces market authorization. Reality: Sandboxes are testing environments, not approval pathways. Manufacturers still need to pursue standard market authorization (CE marking, FDA clearance/approval) based on the evidence generated.

Myth: Only startups benefit from sandboxes. Reality: Large manufacturers with AI-enabled features in existing devices can use sandboxes to validate regulatory strategies for adaptive algorithms and evolving software.

Myth: Sandbox findings are proprietary. Reality: Most programs require transparency and publication of findings — this is how regulators build the evidence base for future framework development.

What's Coming Next

2026 Timeline

Predicted Developments

• More regulators will launch sandbox programs as AI medical devices proliferate. Australia, Japan, and South Korea are the most likely next movers.
• International sandbox collaboration will increase. The MHRA has already signaled interest in cross-border sandbox activities, and the Nature-published analysis of global sandboxes (npj Digital Medicine, 2026) identified this as a key trend.
• Sandbox findings will shape formal regulation. The MHRA's PCCP guidance, the EU MDR/IVDR revisions, and the FDA's evolving AI framework are all being informed by what regulators learn in sandbox environments.
• Enforcement discretion will become a formal tool. TEMPO's approach — conditional enforcement discretion with real-world data collection — may become a standard regulatory pathway for digital health devices.

Key Takeaways

1. Regulatory sandboxes are now a permanent feature of the medical device landscape. The MHRA's multi-year funding and the FDA's TEMPO program demonstrate that these are not pilots — they are infrastructure.

2. AI medical device manufacturers should actively engage with sandboxes. The evidence generated and relationships built during sandbox participation directly de-risk future market authorization.

3. TEMPO is the most innovative program globally because it links regulatory access with payment reform, addressing the two biggest barriers to digital health adoption simultaneously.

4. The EU is behind but catching up. The MDR/IVDR targeted revisions include sandbox provisions, and the EU's formal approach will likely launch in 2027–2028.

5. All manufacturers should monitor sandbox developments, not just AI companies. The regulatory frameworks emerging from these programs will define expectations for the entire industry.