Indigenous Data Governance Offers a Model for AI and Genomics Accountability, Tsosie Argues at AAAS

Uranium mining on Navajo Nation left a health legacy that persists across generations. Contaminated water sources, abandoned mines leaching radioactive dust, and elevated cancer rates in communities near processing sites are documented facts of recent American history. What is less often acknowledged, Krystal Tsosie argues, is how science contributed to that harm - not only by failing to prevent it, but by subsequently interpreting the resulting health problems through a genetic lens that shifted attention away from the environmental causes and the question of accountability.

Tsosie, an assistant professor in the School of Life Sciences at Arizona State University and a leading figure in Indigenous genomics and data governance, brought that history to the American Association for the Advancement of Science Annual Meeting in a talk titled "The Future of Science Is Indigenous." Her argument was not nostalgic or merely critical. It was structural: the governance principles embedded in Indigenous ways of knowing - multigenerational accountability, community consent, reciprocal stewardship of data and land - address specific failure modes in how science is currently organized around genomics, artificial intelligence, and large-scale data infrastructure.

Genomics Is Governance

Tsosie's framing of genomics as fundamentally a governance question rather than a purely technical one runs against how the field often presents itself. Genome sequencing produces data - specific, verifiable, reproducible. What happens to that data, who owns it, who benefits from the research it enables, and how communities are protected if findings are misused: these are governance decisions, not technical ones, and they have profound long-term consequences.

For Indigenous communities, those consequences are not hypothetical. Historical genetics research in Indigenous populations has, in documented cases, been conducted without meaningful community consent, with data used for purposes beyond what participants agreed to, and with findings deployed in ways that conflicted with community interests - including challenges to land rights based on genetic ancestry arguments. The harms are concrete and have reinforced justified skepticism toward research institutions.

"Science has always claimed to study the future. Indigenous peoples have always planned for it," Tsosie said. "We are at a turning point in genomics, AI and precision health. The question is not what we can build, but who science is built for."

What Indigenous Governance Frameworks Actually Look Like

Tsosie is careful to present Indigenous science as a framework rather than a cultural perspective - a distinction that matters for how institutions might engage with it. Indigenous governance models are not monolithic; they vary across nations and traditions. But they share structural features that she argues have direct application to current technology governance challenges.

Seven-generation thinking - the principle that decisions should account for consequences extending to descendants seven generations forward - is one such feature. It is not a metaphor; it is a planning methodology that forces decision-makers to articulate long-term consequences explicitly. In the context of genomic data, which can be linked back to individuals, families, and communities indefinitely as sequencing technology advances, thinking beyond a five-year research horizon is not optional - it is ethically necessary.

Reciprocity is another structural feature: the idea that research relationships involve obligations in both directions, not just extraction from studied communities in service of institutional and academic interests. What does the community get from the research? Who controls the findings? How are harms addressed? These questions, standard in Indigenous research ethics frameworks like the CARE Principles for Indigenous Data Governance, are not routinely built into standard institutional review board processes.

AI Infrastructure and the Extraction Question

Tsosie connected these historical and genomic governance arguments to two current technology expansion trends that she argued carry analogous risks. The first is data center expansion into water-scarce regions like Arizona - where large language model training and inference operations require substantial water for cooling. Indigenous communities in the Southwest have watched other extractive industries deplete water resources before. Whether AI infrastructure will be planned with genuine community input and binding accountability to local water systems, or whether it will repeat that pattern, is an open question that she argued science institutions should be addressing now.

The second is renewed genomic research in Indigenous populations, driven by precision medicine's recognition that current genomic databases are heavily skewed toward populations of European descent, limiting the applicability of polygenic risk scores and pharmacogenomic findings to other populations. The scientific rationale for expanding genomic diversity is genuine - but Tsosie argued that scientific necessity does not override the governance requirements that past failures made clear. Broadening genomic data collection requires building the consent infrastructure, data sovereignty frameworks, and community benefit agreements that earlier research waves did not include.

An Institutional Question

Tsosie's core argument at AAAS was ultimately directed at institutions - universities, funding agencies, journals, professional societies - rather than at individual researchers. Individual researchers often recognize governance obligations; the question is whether the institutional structures they work within reward or penalize acting on those obligations. Grant timelines that prioritize rapid data collection over community relationship-building, publication norms that do not require community co-authorship for research in their territories, and intellectual property frameworks that vest data ownership in universities rather than source communities - these structural features drive behavior independent of individual intent.

"Indigenous science is not new. What is new is whether institutions are willing to recognize Indigenous ways of knowing and include them in advancing science responsibly," Tsosie said.