Artificial intelligence development risks repeating patterns of cultural extraction that have harmed Indigenous peoples for centuries, researchers warn as Australia observes the 50th anniversary of NAIDOC Week, which celebrates the world's oldest continuous living culture.
The core issue centers on a principle long advocated by Indigenous communities: "nothing about us, without us." AI systems trained on datasets often incorporate Indigenous knowledge, cultural practices, and traditional ecological information without obtaining consent, providing attribution, or ensuring benefits flow back to source communities.
This concern reflects legitimate historical precedent. Technologies from agriculture to medicine have extracted Indigenous knowledge through colonization and institutional power imbalances. AI training datasets frequently include digitized cultural materials, ecological data, and traditional practices harvested from public sources without Indigenous oversight or compensation.
The integration of Indigenous knowledge into AI development requires fundamentally different approaches. Rather than treating Indigenous knowledge as raw material for extraction, developers must establish genuine partnerships with Indigenous communities. This means obtaining informed consent before incorporating any cultural information, crediting Indigenous sources explicitly, and structuring agreements so communities share in benefits generated by AI systems using their knowledge.
Some frameworks already exist. The Local Contexts Traditional Knowledge (TK) Labels provide digital markers for cultural materials, signaling restrictions on use and commercial applications. Indigenous data sovereignty movements push for community control over information collection and application. These tools remain underutilized in mainstream AI development.
The timing matters. As AI systems increasingly influence resource management, environmental policy, and healthcare decisions, excluding Indigenous perspectives particularly in ecological and health domains represents a missed opportunity and an ethical failure. Indigenous ecological management practices accumulated over tens of thousands of years contain insights that could improve AI applications in environmental monitoring and conservation.
Building AI with Indigenous knowledge rather than against it requires structural change. Developers must hire Indigenous staff, fund Indigenous-led research, establish community advisory boards with real decision-making authority, and create revenue-sharing arrangements. This represents not charity but recognition