New Delhi / Online — December 3, 2025. The Datasphere Initiative, UNDP, and Kalpa Impact convened an international dialogue on how countries can design AI-enabled Digital Public Infrastructure (DPI) that is safe, inclusive, and aligned with public needs.
As an official Pre-Summit event of the India AI Impact Summit 2026, “Sandboxing: Prototyping and experimentation lessons for AI-powered DPI” brought together policymakers, technologists, and civil-society leaders to examine how real-world experimentation can help governments identify risks and adapt governance models before scaling national digital systems.
How countries are preparing AI-ready Digital Public Infrastructure
Opening the event, Kavita Bhatia (Ministry of Electronics and IT, Government of India), situated the conversation within India’s broader ambition to lead globally on inclusive, development-oriented AI. Drawing on examples such as Aadhaar and UPI, she emphasized that as AI becomes integrated into these large-scale systems, countries will require structured mechanisms like sandboxes to test, validate, and govern new technologies.
Alex Oprunenco (UNDP) highlighted the importance of experimentation, particularly in the Global South, where sudden or rigid technology deployments can have outsized consequences. Sandboxes, he noted, reduce uncertainty, build governance capacity, and allow governments to embed inclusion in new digital services.
Why sandboxes are becoming essential for AI and DPI
Moderated by Sophie Tomlinson (Datasphere Initiative), the first session showcased practical insights from countries already deploying sandboxes at scale.
Narun Popattanachai (Office of the Council of State, Thailand) explained that Thailand has run more than 30 sandbox projects. A failure rate of 30–40%, he noted, is a positive sign: it indicates that risks and limitations are being uncovered early, rather than after national deployment.
From the Dominican Republic, Armando Manzueta (Ministry of Economy) described sandboxes as one of the most effective tools for their national digital strategy, enabling the country to move beyond sector-specific tests toward a coordinated national sandbox strategy.
In Tanzania, Dr. Moses Nkundwe (Jamii Stack Sandbox) underlined how sandboxes support compliance with new data protection laws and help developers safely build on the Jamii DPI stack. In India, Ramesh MC (Center for Open Societal Systems) pointed to India’s own DPI journey, which evolved through repeated cycles of prototyping rather than single “big-bang” launches.
Technical perspectives reinforced this message. Jonathan Middleton (NayaOne) noted that sandboxes enable collaborative development of API standards before public rollout, reducing integration risks. From Brazil, Thiago Moraes (Datasphere Initiative) shared how the Central Bank used sandboxing to test and refine the PIX payment infrastructure ahead of national deployment.
Managing risks in AI-enabled DPI through structured testing
The second session, moderated by Sushant Kumar (Kalpa Impact), examined governance challenges emerging at the intersection of AI and DPI.
Chrissy Meier (DIAL) highlighted opportunities for AI to expand accessibility, from real-time translation to improved service delivery for local innovators. Sourav Das (Co-Develop) stressed that political, institutional, and sovereignty issues continue to shape what is possible, urging governments to adopt an “infrastructure mindset” grounded in clear use cases.
Astha Kapoor (Aapti Institute) cautioned that models trained on skewed DPI datasets risk amplifying existing biases. She encouraged teams to incorporate adversarial thinking into sandbox design, asking questions such as how a criminal or bad actor might exploit a system. Risper Onyango (Africa Sandboxes Forum) emphasized that trust hinges on deeper civil-society involvement, noting that many organizations currently encounter only finished products rather than the design and governance process.
Verena Kontschieder (Prototype Fund CH) argued that responsible innovation requires multiple experimentation spaces, giving early-stage innovators a safe path to collaborate with public institutions before technical decisions become locked in.
Next steps: co-creating AI governance solutions at the India AI Impact Summit
Closing the event, Mariana Rozo-Paz (Datasphere Initiative) introduced the “TrustStack” concept, which brings infrastructure thinking to increase trust in digital systems and to ensure that DPI and AI-powered DPI innovations create value for communities through co-creation with people, iterative testing and intentional transparency.
Insights from this pre-summit dialogue will directly inform the DPI Sandbox Co-Creation Lab, to be hosted by the Datasphere Initiative, UNDP, and Kalpa Impact at the India AI Impact Summit in February 2026.
About the Hosts
The Datasphere Initiative is a global think-and-do-tank shaping the future of data governance by connecting policymakers, innovators, researchers, and civil society to co-create solutions for responsible data use.
UNDP supports countries in leveraging AI and DPI responsibly to advance national development priorities.
Kalpa Impact partners on high-impact digital transformation projects, strengthening responsible adoption of emerging technologies.
The post Experts Urge Real-World Testing for AI-Powered Digital Public Infrastructure: Lessons Ahead of the AI Impact Summit 2026 appeared first on The Datasphere Initiative.