As Artificial Intelligence becomes increasingly embedded into the digital foundations of public life, the window to ensure these systems are safe, inclusive, and accountable is narrowing. To address this, the Datasphere Initiative, UNDP, and Kalpa Impact convened a high-level session at the India AI Impact Summit, moving the global conversation from theoretical principles to the practical realities of implementation.
Building on the official Pre-Summit dialogue held in December 2025, the session titled “AI-DPI experimentation and the role of Sandboxes” gathered senior government officials, development experts, and civil society leaders to examine a critical thesis: that responsible Digital Public Infrastructure (DPI) cannot be built through rigid planning alone, but requires structured, iterative testing environments.
Experimentation is no longer a luxury
Opening the session, Kavita Bhatia (Scientist G – Ministry of Electronics and Information Technology, Government of India; COO – IndiaAI Mission) grounded the discussion in India’s decade-long leadership in developing population-scale DPI, such as Aadhaar and UPI. She posited that while AI offers “transformational possibilities” from proactive health surveillance to precision agriculture, it also introduces systemic risks regarding bias and opacity that traditional governance frameworks may struggle to catch.
“The future of AI-powered DPI must not be accidental. It must be deliberate, principled, and collaborative” – Kavita Bhatia
Bhatia argued that in this context, experimentation is “not a luxury, it is a governance necessity,” serving as the primary mechanism to anticipate risks and create feedback loops before systems are deployed at scale.
Launching the report “Sandboxes for DPI”
During the session, Lorrayne Porciuncula (Executive Director, Datasphere Initiative) officially launched the flagship report, “Sandboxes for DPI: Co-creating the blocks of digital trust.
“Sandboxes can be what we call Laboratories for Trust. When done responsibly, they can serve not only as technical testing grounds but as levers for trust, by testing solutions and preventing “silent failures” where technically functional systems quietly harm human rights or exclude marginalized groups.” – Lorrayne Porciuncula
Porciuncula warned that because DPI carries higher stakes than sector-specific applications, technical flaws can unintentionally “lock in” exclusion for millions. The report, which provides the first global mapping of 16 sandbox initiatives, proposes a shift toward testing environments where governments can test not only the technical code of a system but also its interaction with human rights and legal frameworks to prevent “silent failures”.
Designing systems for people instead of just government efficiency
The panel discussion, moderated by Sushant Kumar (Founder and CEO, Kalpa Impact), moved beyond technical definitions to confront the human reality of digital governance. A recurring theme was the danger of deploying technology without a clear problem statement.
“Digital transformation progresses at the speed of trust. If citizens, governments, civil society do not coordinate and there’s no trust, digital transformation could stall or even be a failure.” – Sushant Kumar
Alexandru Oprunenco (Team Leader, Innovation and Digital, Asia-Pacific Regional Bureau for Asia and the Pacific, UNDP) cautioned against the “blank face” phenomenon, where governments deploy AI solutions without being able to articulate the specific use case or beneficiary. He urged policymakers to constantly ask, “For whom is this system built?”, noting that without this clarity, technology tends to solve government efficiency problems rather than citizens’ service delivery needs.
“The second question which is really important is: for whom? DPI and AI is for whom? […] Is it so that people in the government feel better about what they do and need to spend less time on something, or is it about the people for whom the services are delivered?” – Alexandru Oprunenco
Framing the core tension of the debate, Kumar reminded the room that “digital transformation progresses at the speed of trust.” Citing the example of the “Jamii Number” (Society Number) in Tanzania, he highlighted that while foundational ID and payment stacks are designed to formalize economies, they rely entirely on the public’s confidence. If citizens, governments, and civil society do not coordinate, even the most technically robust systems will stall or fail to be adopted.
Building on this, Dr. Nkundwe Moses Mwasaga (Director General, Information and Communication Technologies Commission, Tanzania) outlined the structural realities of preparing a nation of 62 million people for AI integration. He emphasized that the barrier to success is often not the technology itself, but the human capacity to use it. Dr. Mwasaga identified “digital skills” and “digital security” as the primary challenges, stressing that governments must prioritize consumer protection and privacy frameworks so that citizens, many of whom are not computer savvy, feel safe interacting with these new platforms.
“Our population is 62 million, and there are people who are not computer savvy… Providing basic digital skills to everyone is one of the challenges. Trust is protecting personal data for everyone, the issue of privacy, and also the issue that has to do with consumer protection.” – Dr. Nkundwe Moses Mwasaga
Adding to this perspective, Adesh Khadka (Joint Secretary, Ministry of Communication and Information Technology, Government of Nepal) highlighted the tension between promoting innovation and maintaining regulatory safeguards. He noted that while Nepal’s new AI policy incorporates sandboxing to bridge this gap, practical success requires a cultural shift. Governments must build the technical capacity of regulators and deliberately foster an institutional “acceptance of failure”. Furthermore, Khadka emphasized that civil servants are critical stakeholders whose internal incentives must be navigated to successfully operationalize new technologies.
“How do we improve the capability and capacity at the regulators… and at the same time, how do we introduce this acceptance of failure? Because these are experiments at the end, and there could be a failure… how do we make sure that the traditional regulator has this vision that this might fail and how do we accept it?” – Adesh Khadka
Why we must test AI in the context of real public infrastructure
Bridging the gap between policy and code, Morine Amutorine (Africa Sandboxes Forum Lead, Datasphere Initiative) and Dr. Verena Kontschieder (Co-CEO Opendata.ch; Program Lead Prototype Fund Switzerland) argued that sandboxes must evolve from tools for market entry into mechanisms for societal oversight.
“We see the intersection between AI and DPI as an opportunity to not just test AI in isolation, but to test it in a context that is already set up for systems that are built to serve greater people.” – Morine Amutorine
Amutorine noted that sandboxes offer the unique ability to “test AI in context,” allowing authorities to observe how algorithms behave when embedded in real infrastructure and interacting with diverse legal systems. Kontschieder added that this testing must go beyond technical robustness to ensure “healthy societal considerations,” effectively preventing functional systems from quietly harming marginalized groups.
“Sandboxes are sometimes really used to facilitate access to market instead of trying to work out a balance of healthy societal considerations, and healthy economic considerations, and technological considerations.” – Verena Kontschieder
Turning these insights into a toolkit for action
The session concluded with a consensus that the next phase of DPI governance requires moving from ad hoc pilots to institutionalized capacity for testing.
To support this transition, the Datasphere Initiative announced that the findings from the new report will be translated into a practical “How-to Toolkit” throughout 2026, alongside a dedicated track at the upcoming Sandbox Summer School in Lisbon.
Read the full report: Sandboxes for DPI: Co-creating the blocks of digital trust
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 At India AI Impact Summit, global authorities establish experimentation as a governance necessity for responsible digital public infrastructure appeared first on The Datasphere Initiative.