Following the rich discussion captured in our earlier post, this second piece distills the core insights, challenges, and recommendations that emerged from the Official Pre-Summit event “Sandboxing: Prototyping and experimentation lessons for AI-powered DPI”. These reflections will directly feed into the DPI Sandbox Co-Creation Lab taking place at the India AI Impact Summit in February 2026.
Emerging themes for safe and inclusive AI-powered DPI
The session brought together practitioners, policymakers, researchers, and civil-society leaders from India, Brazil, Colombia, the Dominican Republic, Kenya, Switzerland, Thailand Tanzania, and the United Kingdom, along with global institutions, to explore how countries can responsibly design, test and govern AI-powered Digital Public Infrastructure (DPI). Several clear themes emerged:
1. Experimentation and iteration must become the norm
Speakers stressed that AI-enabled DPI cannot rely on linear or top-down deployments. Countries need structured mechanisms, such as sandboxes, staged testing environments, synthetic datasets, and modular pilots, to uncover risks early while ensuring scalability and safety.
2. Trust-building is foundational
The rise of AI-driven scams, misinformation, and opaque automated decisions has weakened public confidence. Participants emphasized the importance of user-centered design, youth engagement, transparency about system behavior and collaborative spaces where assumptions can be tested openly.
3. Transparency, accountability and active monitoring are urgent gaps
Harms such as exclusion, biased verification, and data misuse often remain undocumented. Concrete safeguards identified during the session included documenting data flows, conducting targeted audits and running stress tests to understand system behavior under pressure.
4. AI and DPI must be understood as interconnected infrastructures
AI strengthens DPI through multilingual capabilities, personalization, and accessibility, while DPI provides the foundations and the data governance, interoperability, and public trust that ensure AI is deployed responsibly. Participants underscored the need for sustainable implementation models and improved interoperability between DPI and AI.
5. Meaningful inclusion must be designed
Persistent risks of exclusion affect women, youth, rural communities, Indigenous peoples, and other marginalized groups. Co-creation processes and sandboxes were highlighted as practical pathways to embed lived experiences, strengthen digital literacy, and align systems with real-world needs.
South–South collaboration and sustained capacity-building emerged as cross-cutting priorities. Countries across the Global South are already pioneering governance models, and scaling these efforts will require long-term investment in institutions, skills, and civil society engagement.
Looking ahead: co-creating solutions at the India AI Impact Summit
As we move toward the India AI Impact Summit 2026, these insights will shape a co-creation lab, jointly hosted by the Datasphere Initiative, UNDP and Kalpa Impact. The Lab will provide a space for governments, innovators, and civil society to collaboratively develop solutions to the governance challenges surfaced during this Pre-Summit dialogue.
If you want to contribute to the conversation on how prototyping, testing and sandboxing can support AI-powered DPI, fill out this form.
Together, these conversations mark an important step toward building AI-powered public infrastructure that is trustworthy, inclusive and grounded in real-world needs.
