Who Governs the Data? Community-Led AI from the Global South

Data governance is one of the most discussed topics in AI policy.

The panel hosted by the NYU Peace Research and Education Program at the AI & Equality Festival of Ideas brought something different to that conversation: practitioners who are building community authority over data in real places, with real communities, under real constraints. Moderated by Marine Collins Ragnet, the session gathered Angela Oduor Lungati from Ushahidi, Mark Irura from the UNDP Digital AI and Innovation Hub in Nairobi, and Tala Alngag Bautista from the indigenous Kolinga community and Ifugao State University in the Philippines. The conversation moved through three questions: what community data governance looks like when it is built from the ground up, what holds that authority in place when funding shifts or political conditions change, and who is still missing from these rooms.

Angela Oduor Lungati opened with a reframing that set the tone for everything that followed. For most of Ushahidi’s nearly two decades of work collecting community-generated information across elections, crises, and movements, the value of community data was understood through the lens of where it came from and who generated it. In the AI era, that meaning has shifted toward ownership. She referenced a phrase shared by Dr. Joy Buolamwini during a Coded Bias event co-hosted by Ushahidi in Nairobi: data is destiny. The question for communities today is no longer who contributed the data, but who gets to shape the future that data produces.

Mark Irura built on that by walking through what community authority over data looks like in practice. He drew a distinction between the right to consent and the right to say no, and argued that the latter remains largely unresolved. His work on the Kenya Language Corpus Framework and his involvement in Kenya’s national AI strategy point toward practical architectures: layered consent models that allow communities to withdraw consent at different stages of a data pipeline, custodianship frameworks grounded in existing legislation such as the Kenya Culture Preservation Act, and data cooperative models that give communities economic stake in the infrastructure they help build. He pointed to the Noodle License as one of the few existing instruments that practically encodes community benefit-sharing into the terms of data use.

Tala Alngag Bautista highlighted  what it looks like when AI governance begins from indigenous values rather than adapting them in later. The Kolinga community in the Philippines is building a provincial AI ordinance rooted in four indigenous values: paneo, governing the relationship with the spiritual world; nilin, governing the relationship with the self and with creation; bain, governing relationships with others; and atom, a principle of self-restraint meaning that something should not be used if it causes harm. When Marine Collins Ragnet and her colleagues from NYU Peace explained the environmental cost of AI to Kolinga elders, there was a long pause, and then someone said: so we shouldn’t use AI at all. That pause, Tala said, carries the full weight of a value system built around interconnection and the responsibility not to harm others, including unknown communities whose water is consumed by data centers on the other side of the world.

Marine Collins Ragnet grounded the whole panel in a peacebuilding frame: AI is a system capable of perpetrating structural violence through the power asymmetries and patterns of extraction it enables. The work of community-centered data governance is therefore peacebuilding work. Her research across Malawi, Kenya, and the Philippines through NYU Peace AI points to three recurring challenges: funding cycles that are too short for the trust-building that genuine community participation requires, the gap between community governance frameworks and binding national law, and the cost of building technical infrastructure that communities actually control. Ongoing work includes a data commons for Ushahidi’s nineteen years of community intelligence in partnership with Oxford’s Institute for Ethics in AI, a data collaborative for Kolinga indigenous knowledge in the Philippines, and capacity-building workshops with data labelers in Kenya.

The panel closed on a question about who is still missing: local communities are present in these conversations as case studies and data points, and need to be present as decision makers who hold real authority over the systems that shape their lives.

Getting there requires redistributing power practically, not just normatively, through technical infrastructure, legal frameworks, and funding models that treat community governance as a structural commitment rather than a design principle.